distribution-free model for ambulance location...
TRANSCRIPT
Research ArticleDistribution-Free Model for Ambulance LocationProblem with Ambiguous Demand
Feng Chu1 Lu Wang2 Xin Liu 3 Chengbin Chu34 and Yang Sui5
1Management Engineering Research Center Xihua University Chengdu 610039 China2Department of Aviation Transportation Shanghai Civil Aviation College Shanghai 200232 China3School of Economics amp Management Tongji University Shanghai 200092 China4Laboratoire Genie Industriel Centrale Supelec Universite Paris-Saclay Grande Voie des Vignes Chatenay-Malabry France5Glorious Sun School of Business amp Management Donghua University Shanghai 200051 China
Correspondence should be addressed to Xin Liu liuxin9735126com
Received 11 June 2018 Accepted 28 August 2018 Published 13 September 2018
Academic Editor Oded Cats
Copyright copy 2018 FengChu et alThis is an open access article distributed under the Creative CommonsAttribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
Ambulance location problem is a key issue in Emergency Medical Service (EMS) system which is to determine where to locateambulances such that the emergency calls can be responded efficiently Most related researches focus on deterministic problems orassume that the probability distribution of demand can be estimated In practice however it is difficult to obtain perfect informationon probability distributionThis paper investigates the ambulance location problem with partial demand information ie only themean and covariance matrix of the demands are known The problem consists of determining base locations and the employmentof ambulances to minimize the total cost A new distribution-free chance constrained model is proposedThen two approximatedmixed integer programming (MIP) formulations are developed to solve it Finally numerical experiments on benchmarks (Nickel etal 2016) and 120 randomly generated instances are conducted and computational results show that our proposed two formulationscan ensure a high service level in a short time Specifically the second formulation takes less cost while guaranteeing an appropriateservice level
1 Introduction
Thedesign of EmergencyMedical Service (EMS) systems thataffects peoplersquos health and life focuses on how to respond toemergencies rapidly Ambulance location problem is one ofthe key problems in EMS system which mainly consists ofdetermining where to locate bases also named as emergencyservice facilities and the employment of ambulances in orderto serve emergencies efficiently and to guarantee patient sur-vivability There have been various researches investigatingambulance location problem (eg [1 2]) since it is introducedby Toregas et al (1974)
Early studies addressing ambulance location problemmainly focus on deterministic environments including setcovering ambulance location problem that minimizes thenumber of ambulances to cover all demand points [3]maximal covering location problem tomaximize the numberof covered demand points with given number of ambulances
[4] and double standard model (DSM) in which eachdemand point must be covered by one or more ambulances[5] However in practice stochastic ambulance locationproblem is more realistic due to the inevitable uncertaintiesof emergency events (eg [6 7] etc) Usually an emergencyevent has the following characteristics (i) it is difficult toforecast precisely where an emergency will occur and (ii) therequired number of ambulances depends on the severity ofthe situation which is also unforeseen [8]
Most existing works investigate stochastic ambulancelocation problem by assuming that the demand probabilitydistribution at each possible emergency is known (eg [910] Beraldi and Bruni 2008 etc) However as stated byWagner [11] and Delage and Ye [12] it is usually impossibleto obtain the perfect information on probability distributiondue to (i) the lack of historical data and the fact that (ii) thegiven historical data may not be represented by probabilitydistribution
HindawiJournal of Advanced TransportationVolume 2018 Article ID 9364129 12 pageshttpsdoiorg10115520189364129
2 Journal of Advanced Transportation
Moreover how to cover as many emergencies as possibleto guarantee a high patient service level (ie the portion ofthe satisfied emergency points among all emergency points)has always been a main goal of EMS operators Motivatedby the above observation this work focuses on a stochasticambulance location problem with only partial informationof demand points ie the mean and covariance matrix ofthe demands Besides we introduce a new individual chanceconstraint which implies a minimum probability that eachemergency demand has to be satisfied For the problema new distribution-free model is presented and then twoapproximated formulations are proposed The contributionof this work mainly includes the following
(1) For the studied stochastic ambulance location prob-lem we introduce a new individual chance con-straint (ie safety level) guaranteeing each emergencydemand satisfied with a least probability while Nickelet al [8] use a coverage constraint where some serviceof individual emergencies may suffer insufficiency
(2) A new distribution-free model based on individualchance constraints is proposed To our best knowl-edge it is the first distribution-freemodel for stochas-tic ambulance location problem
(3) To solve the distribution-free model two approx-imated mixed integer programming (MIP) formu-lations are developed Experimental results showthat our approximated formulations are efficient andeffective for large size instances compared to thatproposed by Nickel et al [8] in terms of both com-putational time and service level
The remainder of this paper is organized as followsSection 2 gives a brief literature review In Section 3 we givethe problem description and propose a new distribution-freemodel In Section 4 two approximated MIP formulationsare proposed Computational results on benchmarks and120 randomly generated instances are reported in Section 5Section 6 summarizes this work and states future researchdirections
2 Literature Review
The deterministic ambulance location problem has been wellstudied in literature (eg [2 5 13] etc) Besides there havebeen some works addressing ambulance location problemunder uncertainty (eg [1 6 14] etc) Since our study fallswithin the scope of stochastic ambulance location problem inthe following subsections we first review existing studies onambulance location problem with uncertain demand Thenwe review the literature studying general and specific stochas-tic optimization problems with distribution-free approaches
21 Ambulance Location Problem with Uncertain DemandAmbulance location problem under uncertain demand hasbeen investigated by many researchers Most existing worksaddress the uncertainty with given scenarios or knownprobability distribution
Chapman and White [15] first investigate the ambulancelocation problem with uncertain demand in which the
complete information on probability distributions is assumedto be known Beraldi et al [9] study the emergency medicalservice location problemwith uncertain demand tominimizethe total cost with amarginal probability distribution Beraldiand Bruni (2008) investigate the emergency medical servicefacilities location problem with stochastic demand based ongiven set of scenarios and known probability distributionA stochastic programming formulation with probabilisticconstraints is proposed Noyan [10] studies the ambulancelocation problem with uncertain demand on given scenariosand probability distribution Then two stochastic optimiza-tion models and a heuristic are proposed
Recently Nickel et al [8] first study the joint optimizationof ambulance base location and ambulance employmentTheobjective is to minimize the total constructing base costsand ambulance employment costs It is assumed that variousscenarios are given beforehand A coverage constraint isintroduced in the paper They propose a sampling approachthat solves a finite number of scenario samples to obtaina feasible solution of the original problem But with thecoverage constraint some emergency demands risk insuffi-cient individual service or cannot be served For the studywe propose a new distribution-free model for stochasticambulance location problem
22 Distribution-Free Approaches In data-driven settingsthe probability distributions of uncertain parameters maynot always be perfectly estimated [16] Therefore in the lastdecade there have beenmany solution approaches developedto address stochastic problems under partial distributionalknowledgeMost related researches focus on the distribution-free approach via considering chance constraintsWagner [11]studies a stochastic 0-1 linear programming under partialdistribution information ie min119883isin01119899c1015840x a1015840119895x le119887119895 119895 = 1 119898 where a119895 are random vectors withunknown distributions The only information on a119895 aretheir moments up to order 119896 A robust formulation asa function of 119896 is given Given the known second-ordermoment knowledge ie 119896 = 2 an approximated formulationis developed as min119883isin01119899c1015840x radicx1015840Γ119895x le radic119901119895(1 minus 119901119895)(119887119895 =E[a119895]1015840x) 119895 = 1 119898 where Γ119895
119894119896= E[(119886119894119895 minus E[119886119894119895])(119886119896119895 minus
E[119886119896119895])] forall119894 119896 = 1 119899 Delage and Ye [12] investigate thestochastic program with limited distribution informationand they propose a new moment-based ambiguity set whichis assumed to include the true probability distribution todescribe the uncertainties There have been various workssuccessfully applying the distribution-free approaches Ng[17] investigates a stochastic vessel deployment problem forliner shipping in which only the mean standard deviationand an upper bound of demand are known A distribution-free optimization formulation is proposed Based on that Ng[18] studies a stochastic vessel deployment problem for theliner shipping industry where only the mean and varianceof the uncertain demands are known New models areproposed and the provided bounds are shown to be sharpunder uncertain environment The stochastic dependenciesbetween the shipping demands are considered
Jiang and Guan [19] develop approaches to solve stochas-tic programs with data-driven chance constraints Two types
Journal of Advanced Transportation 3
of confidence sets for the possible probability distributionsare proposed For more distribution-free formulation appli-cations please see Lee and Hsu [20] Kwon and Cheong [21]etc for the stochastic inventory problem and Zhang et al[23] for stochastic allocating surgeries problem in operatingrooms and Zheng et al [24] for stochastic disassembly linebalancing problem To the best of our knowledge there is noresearch for the stochastic ambulance location problem withonly partial information on the uncertain demand
3 Problem Description and Formulation
In this section we first describe the considered problem andthen propose a new distribution-free model
31 Problem Description There is a given set of candidatebase locations 119868 = 1 2 |119868| and a set of potentialemergency demand points 119869 = 1 2 |119869| The emergencypoints may refer to a road a part of the urban area and avillage Base 119894 isin 119868 is said to be covering an emergency point119895 if the driving time 119905119894119895 between 119894 and 119895 is no more than apredetermined value119879The set 119868119895 of candidate bases coveringemergency point 119895 is denoted as 119868119895 = 119894 isin 119868 | 119905119894119895 le 119879
The number of ambulances required by an emergencypoint 119895 isin 119869 is denoted as 119889119895 which depends on the severityof the practical situationWe consider uncertain emergencieswhich are estimated or predicted by partial distributionalinformation thus the demand is regarded to be ambiguousBesides we focus on the case where the historical data cannotbe represented by a precise probability distribution It isassumed that the customer demands are independent of eachother
The objective of the problem is to select a subset ofbase locations and determine the number of ambulances ateach constructed base in order to minimize the total baseconstruction cost and ambulance costThroughout this paperwe assume that only the mean and covariance matrix ofdemands are known Moreover a predefined safety level 120572119895is given for each potential emergency point 119895 isin 119869 That is thenumber of ambulances serving each emergency point 119895 isin 119869 islarger than or equal to its demand with a least probability of12057211989532 Chance Constraint Construction In this section weintroduce a chance constraint to guarantee the safety levelof each emergency demand point In the following 119910119894119895 is adecision variable denoting the number of ambulances servingpoint 119895 isin 119869 from base 119894 isin 119868119895 and 119889119895 is the uncertain demandat point 119895 isin 119869 The chance constrained inequality is presentedas follows
ProbP (sum119894isin119868119895
119910119894119895 ge 119889119895) ge 120572119895 forall119895 isin 119869 (1)
where ProbP(sdot) denotes the probability of the event inparentheses under any potential probability distribution PConstraint (1) ensures that the number of ambulances servingemergency point 119895 isin 119869 is no less than its demand with a leastprobability of 120572119895 (ie safety level)
33 Distribution-Free Formulation In the following we givebasic notations define decision variables and propose thedistribution-free formulationDF for the ambulance locationproblem with uncertain demand
Parameters
119894 index of base locations119895 index of emergency points119868 set of candidate base locations119869 set of possible emergency demand points119905119894119895 driving time between base 119894 isin 119868 and demand point119895 isin 119869119879 maximum driving time for serving any emergencycall119868119895 set of candidate bases that can cover emergencypoint 119895 isin 119869 ie 119868119895 = 119894 isin 119868 | 119905119894119895 le 119879119869119894 set of potential emergency points covered by base119894 isin 119868 ie 119869119894 = 119895 isin 119869 | 119905119894119895 le 119879119891119894 fixed construction cost for installing a base atlocation 119894 isin 119868119892119894 fixed cost associated with an ambulance to belocated at 119894 isin 119868119889119895 number of (stochastic) ambulances requested byemergency point 119895 isin 119869M a sufficiently large positive number
Decision Variables
119909119894 a binary variable equal to 1 if a base is constructedat 119894 isin 119868 0 otherwise119911119894 number of ambulances assigned to possible baselocation 119894 isin 119868119910119894119895 number of ambulances serving emergency point119895 isin 119869 from possible base location 119894 isin 119868119895
Distribution-Free Model [DF]
[DF]
min sum119894isin119868
(119891119894 sdot 119909119894 + 119892119894 sdot 119911119894)st Constraint (1)
(2)
119911119894 le M sdot 119909119894 forall119894 isin 119868 (3)
sum119895isin119869119894
119910119894119895 le 119911119894 forall119894 isin 119868 (4)
119909119894 isin 0 1 forall119894 isin 119868 (5)
119911119894 isin Z+ forall119894 isin 119868 (6)
119910119894119895 isin Z+ forall119894 isin 119868 119895 isin 119869119894 (7)
4 Journal of Advanced Transportation
The objective function denotes the goal to minimize thetotal cost consisting of two parts (i) the cost for constructingbases ie sum119894isin119868 119891119894 sdot 119909119894 and (ii) the cost for deploying ambu-lances ie sum119894isin119868 119892119894 sdot 119911119894
Constraint (3) ensures that ambulances can only belocated at the opened bases Constraint (4) ensures that thenumber of ambulances sent to serve emergency points frombase 119894 isin 119868 does not exceed the total number of ambulanceslocated at 119894 isin 119868 Constraints (5)-(7) are the restrictions ondecision variables
4 Solution Approaches
The proposed distribution-free model is difficult to solvewith the commercial software due to chance constraints Inthis section we propose two approximatedMIP formulationsbased on those in Wagner [11] and in Delage and Ye [12]respectively In the following subsections we present the twoapproximated MIP formulations
41 Approximated MIP Formulation MIP-DF1 In this partwe first describe a widely used ambiguity set to describe theuncertainty Then an approximated MIP formulation MIP-DF1 is proposed
Given a set of independent historical data samples 119889119904|119878|119904=1of random vectors of demands where 119878 is the set of sampleindexes the mean vector 120583 and the covariance matrix Σ ofdemands can be estimated as follows
120583 = 1|119878|sum119904isin119878119889119904Σ = 1|119878|sum119904isin119878 (119889119904 minus 120583) (119889119904 minus 120583)⊤
(8)
where (sdot)⊤ implies the transposition of the vector in paren-theses Then an ambiguity set of all probability distributionsof demandsP1(120583 Σ) can be described as the following
P1 (120583 Σ) = P EP [119889] = 120583EP [(119889 minus 120583) (119889 minus 120583)⊤] = Σ (9)
whereP denotes a possible probability distribution satisfyingthe given conditions and EP[sdot] denotes the expected valueof the number in parentheses Then the chance constraint (1)with the ambiguity setP1 can be presented as follows
ProbP (sum119894isin119868119895
119910119894119895 ge 119889119895) ge 120572119895 forall119895 isin 119869 P isin P1 (10)
According to Wagner (2010) and Ng [18] constraint (1) canbe conservatively approximated by the following
sum119894isin119868119895
119910119894119895 ge 120583119895 + 120590119895 sdot radic 120572119895(1 minus 120572119895) forall119895 isin 119869 (11)
where 120590119895 = radicΣ119895119895 and Σ119895119895 denotes the 119895-th element in the 119895-thcolumn of matrix Σ In terms of conservative approximation
all solutions satisfying constraint (11) must satisfy constraint(1) Then we describe the approximated MIP formulationMIP-DF1 toDF combined with constraint (11)
[MIP-DF1]
min sum119894isin119868
(119891119894 sdot 119909119894 + 119892119894 sdot 119911119894)st Constraints (4) - (8) (12)
(12)
MIP-DF1 can be exactly solved by the commercialsoftware such as CPLEX As we can observe from thecomputational results in Section 5 it can obtain a relativelyhigh service level However the system cost is also high Inorder to save the system cost another approximated MIPformulation of DF is then proposed in the next subsection
42 Approximated MIP Formulation MIP-DF2 AmbiguitysetP1 focuses on exactly matching the mean and covariancematrix of uncertain parameters However in practice theremay be considerable estimation errors in the mean andcovariance matrix To take the inevitable estimation errorsinto consideration Delage and Ye [12] introduce a newmoment-based ambiguity set Therefore we in the followingemploy the moment-based ambiguity set
P2 (120583 Σ 1205741 1205742)= P (EP [119889] minus 120583)⊤ Σminus1 (EP [119889] minus 120583) le 1205741
EP [(119889 minus 120583) (119889 minus 120583)⊤] ⪯ 1205742Σ (13)
where 1205741 ge 0 and 1205742 ge 0 are two parameters of ambiguityset P2(120583 Σ 1205741 1205742) with the following assumptions (i) thetrue mean vector of demands is within an ellipsoid of sizeproportion to 1205741 centered at 120583 and (ii) the true covariancematrix of demands is in a positive semidefinite cone boundedby amatrix inequality of 1205742ΣThe ambiguity set describes howlikely the uncertain parameters are to be close to the meanin terms of the correlation [12] Besides under the moment-based ambiguity set various stochastic programs with partialdistributional information have been successfully modeledand solved [22 23 25]
According to the approximation method in Zhang et al[23] chance constraint (1) can be approximated by
radic 11 minus 119886 minus 119887 sdot (1 + radic1 minus 120572119895120572119895 sdot 119887) sdot 120590119895le radic1 minus 120572119895120572119895 sdot (sum
119894isin119868119895
119910119894119895 minus 120583119895) 119895 isin 119869(14)
where
119886 = 1 minus 1205741 + 11205742 minus 1205741 119887 = 12057411205742 minus 1205741
(15)
Journal of Advanced Transportation 5
Table 1 120572 = 095|119868| = |119869| Sampling approach MIP-DF1 MIP-DF2
CT Obj Sel () CT Obj Sel () CT Obj Sel ()1 7 49 23 9626 01 36 10000 01 23 100002 7 51 22 9863 01 35 10000 01 24 99823 6 41 14 9692 01 30 10000 01 14 98504 6 43 13 9713 02 29 10000 01 12 97425 6 57 12 7996 01 30 10000 02 12 100006 8 58 21 9850 01 24 10000 02 27 10000
Average 50 175 9457 01 307 10000 01 187 9929
The second approximated MIP formulationMIP-DF2 toDF is presented as follows
[MIP-DF2]
min sum119894isin119868
(119891119894 sdot 119909119894 + 119892119894 sdot 119911119894)st Constraints (4) - (8) (15)
(16)
Notice that when 1205741 = 0 and 1205742 = 1 inequality (14)reduces to (11) in MIP-DF1 which implies that MIP-DF1 isa special case of MIP-DF2
5 Computational Experiments
In this section the performance of our proposed formula-tions first evaluated and compared to the sampling approachproposed in Nickel et al [8] Specifically given a requiredservice level 120572 we apply our MIP formulations by restrictingthe chance constraint for each emergency point to satisfya safety level 120572119895 = 120572 while we adopt the samplingapproach by restricting the coverage constraint with 120572 Thencomputational results on 120 randomly generated instancesbased on the information on emergencies in Shanghai arereported Our approximated MIP formulations and thesampling approach are solved by coding in MATLAB 2014bcalling CPLEX 126 solver All numerical experiments areconducted on a personal computer with Core I5 and 330GHzprocessor and 8GB RAMunder windows 7 operating systemThe computational time for sampling approach is limited to3600 seconds
51 Out-of-Sample Test FollowingZhang et alrsquos [22]method-ology we examine the out-of-sample performance of solu-tions obtained by our proposed approximated MIP formula-tions and the sampling approach [8] The out-of-sample testincludes three steps
(1) The base locations and the employment of ambu-lances at each base are determined with known meanvector and covariance matrix of demands which isnamed as in-sample test
(2) 1000 scenarios are sampled from the underlying truedistribution which represent realizing the uncertaindemand
(3) Based on the solution obtained in step (1) ambu-lances are assigned to serve emergency points inthe 1000 scenarios to maximize the (out-of-sample)service level
The (out-of-sample) service level is defined as the ratioof the number of emergency points with demand satisfiedto the total number of emergency points in 1000 scenariosIn the following subsections the benchmarks are tested andthen three common distributions are employed ie uniformdistribution Poisson distribution and normal distributionrespectively
52 Benchmark Instances We first test the benchmarks inNickel et al [8] to compare our proposed two approximatedMIP formulationswith the sampling approach InNickel et al[8] they assume that demand nodes are also potential baselocations and the number of ambulances required by eachemergency point is chosen from set 0 1 2with given proba-bility distribution Computational results on benchmarks arereported in Tables 1 and 2
Table 1 reports the computational results with servicelevel requirement120572 = 095 In Table 1 CTObj and Sel denotethe computational time objective value (ie the system cost)and the service level respectively The first column includesthe indexes of instances and the second indicates the numberof nodes It shows that the average service level obtainedby the sampling approach is 9457 which is smaller thanthe required service level ie 85 Specifically the servicelevel of instance 5 obtained by the sampling approach is only7996 However MIP-DF1 and MIP-DF2 can guarantee theservice levels for all instances higher than the requirementConcluding from Table 1 we can observe that (i) the averagecomputational time of sampling approach is about 50 timesas those of the two approximated MIP formulations and (ii)MIP-DF1 and MIP-DF2 improve the average service level byabout 574 and 499 compared with sampling approachhowever (iii) system costs obtained by MIP-DF1 and MIP-DF2 are respectively 7543 and 686 higher than that ofsampling approach
Table 2 reports the computational results with servicelevel requirement 120572 = 099 For instances 2-6 the ser-vice levels obtained by sampling approach cannot meet therequirement ie 99 Besides it shows that the averagecomputational time of sampling approach is 48 which isabout 48 times greater than those of MIP-DF1 andMIP-DF2
6 Journal of Advanced Transportation
Table 2 120572 = 099|119868| = |119869| Sampling approach MIP-DF1 MIP-DF2
CT Obj Sel () CT Obj Sel () CT Obj Sel ()1 7 49 24 9979 01 64 10000 01 25 100002 7 48 23 9856 01 63 10000 01 24 100003 6 51 15 9717 01 55 10000 01 16 100004 6 47 15 9275 01 52 10000 01 13 100005 6 40 14 9000 01 55 10000 01 14 100006 8 52 23 9785 01 45 10000 01 27 10000
Average 48 190 9343 01 557 10000 01 198 10000
Table 3 Solutions to instance 4 in Nickel et al [8] for the given scenario
Sampling approach MIP-DF1 MIP-DF2Emergency points 1 2 3 4 5 6 Emergency points 1 2 3 4 5 6 Emergency points 1 2 3 4 5 6
Demand 1 0 2 0 2 0 Demand 1 0 2 0 2 0 Demand 1 0 2 0 2 0Base locations Ambulance employment Ambulance employment Ambulance employment1 0 0 02 1 1 0 3 1 23 1 1 26 1 2 2 04 0 0 05 0 6 06 2 2 0 4 2
The average service level is 9343 Although the averageservice levels obtained by MIP-DF1 and MIP-DF2 are both100 the average system cost obtained by MIP-DF1 is about3 times larger than that of sampling approach Besides theaverage system cost obtained by MIP-DF2 is 421 higherthan that of sampling approach
Concluding for the benchmarks described in Nickel etal [8] we can observe that (1) MIP-DF1 and MIP-DF2 canbe solved in far less computational time compared with thesampling approach (2) the sampling approach cannot meetthe requirement of service level for all benchmarks whilethe two MIP formulations can (3) the average system cost ofMIP-DF2 is very close to that of sampling approach (4) withthe increase of 120572 the average system costs obtained by MIP-DF2 is getting closer to that obtained by sampling approach
From Tables 1 and 2 we can observe that in spite ofthe lower system cost obtained by the sampling approachthere may exist extreme situations that demands of someemergency points are not satisfied Take one benchmark inNickel et al [8] ie instance 4 where there are 6 nodesA scenario in which demand at each emergency point isgenerated from the given probability distribution is shown inFigure 1 We can observe from Figure 1 that under the givenscenario the demand of each possible emergency point is1 0 2 0 2 0The solutions obtained by sampling approachMIP-DF1 andMIP-DF2 are shown in Table 3 which includesbase locations and ambulance employment obtained by in-sample test and ambulance assignment under given scenario
In Table 3 the first column includes the indexes of poten-tial base locations The second column represents the baselocations and the ambulance employment at each location
>1 = 1
>2 = 0
>3 = 2
>4 = 0
>5 = 2
>6 = 0
1
2 4
3 5
6
Figure 1 Instance 4 in Nickel et al [8] under given scenario
in which a location with 0 ambulance implies that it is notselected for base construction From Table 3 we can observethat the in-sample solutions obtained by sampling approachis to construct bases at locations 2 3 and 6 where the numberof ambulances is 1 1 and 2 respectively Besides undergiven scenario there is one ambulance serving emergencypoint 1 from base location 3 and one ambulance servingemergency point from base location 2 and 2 ambulancesserving emergency point 5 from base location 6 It shows thatthe requirement of ambulances at emergency point 3 is notfully satisfied The insufficient service may cause serious lossof life However we can find that the solutions obtained byMIP-DF1 andMIP-DF2 can provide sufficient ambulances forall emergency points
In sum to guarantee the required service level withless cost MIP-DF2 is recommended for the solutions with
Journal of Advanced Transportation 7
Nickel et al [8]rsquos data set Moreover in the benchmarks itis assumed that all nodes are possible emergency points aswell as potential base locations and the number of requiredambulances is small which is not always consistent withthe situation in Shanghai Therefore we further test the twoapproximated MIP formulations and the sampling approachon randomly generated instances by analysing the informa-tion on emergencies in Shanghai [26]
53 Computational Tests on Randomly Generated InstancesIn the following subsection numerical experiments on 120randomly generated instances are described
531 Test Instances By analysing the information on emer-gencies in Shanghai [26] the mean value 120583119895 of demand 119889119895 ineach emergency point are randomly generated fromadiscreteUniform distribution on interval [5 25] The standard devia-tion 120590119895 is set to be 5 Since the sampling approach proposed inNickel et al [8] requires a sample of scenarios in numericalexperiments scenarios are produced with demands randomlygenerated from Uniform distribution Poisson distributionand Normal distribution consistent with given mean andstandard deviation
According to the instance settings in Erkut and Ingolfsson(2008) response time from a base to an emergency pointis randomly generated from a discrete Uniform distributionon interval [3 30] in units of minutes A base covers anemergency point if the driving time between them is lessthan 10 minutes to ensure the survival probability (Erkut andIngolfsson 2006) For an emergency point in each instanceif there is no base with 10 minutes driving time to it then thebase covering this point is set to be the nearest one The costof constructing a base and that of locating one ambulance areboth set to be 1 as in Nickel et al [8] Besides the number ofemergency points is 3 times that of potential base locations
532 Results and Discussion Different values of requiredservice level 120572 are tested ie 085 09 095 Computationalresults on 120 randomly generated instances are reported inTables 4-6
Table 4 reports the numerical results of instances withdemand generated from Uniform distribution We observethat when 120572 = 085 09 095 the average computationaltimes of the sampling approach are 4482 4379 and 4295seconds respectively However both approximated MIP for-mulations can produce solutions in much faster speeds iewithin 13 seconds on average With the increase of 120572 (i)the out-of-sample average service level is getting higher and(ii) the system costs of solutions obtained by all method aregetting larger and (iii) the system costs of solutions yieldedbyMIP-DF2 is getting closer to those of solutions obtained bythe sampling approach In terms of the average system costthe sampling approach performs the best while MIP-DF1 isthe poorest When 120572 = 085 for example the average systemcost ofMIP-DF2 and the sampling approach are about 3032and 442 less than that of MIP-DF1 respectively When120572 = 095 the above gaps are even larger Concluding fortheUniformdistribution (1) the proposed two approximatedMIP formulations are far more time saving than the sampling
approach (2)MIP-DF1 andMIP-DF2 can improve the servicelevel by about 1907 compared with the sampling approachhowever (3) despite of the highest service levels obtainedby MIP-DF1 the system cost obtained by MIP-DF1 is alsohigh (4) MIP-DF2 can ensure an appropriate service levelwith much less cost than MIP-DF1 and (5) for the situationswith small-scale instances and high service level requirementMIP-DF1 is recommended
Computational results of instances with demand at eachemergency point generated from Poisson distribution arepresented in Table 5 For MIP-DF1 and MIP-DF2 onlymean and variance are involved Therefore with givenmean and standard deviation of the demand at each pointcomputational results are the same for different probabilitydistributions We can obtain from Table 5 that the objectivevalue obtained by the sampling approach increases with theincrease of 120572 Besides when the value of 120572 equals 085the average service level obtained by the sampling approachis about 8143 which is smaller than that of the twoapproximated MIP formulations The highest service level9934 is obtained by MIP-DF1 and that obtained by MIP-DF2 is 9708 When 120572 = 09 the average service levelof the sampling approach is 8437 and those obtained byMIP-DF1 and MIP-DF2 are 9985 and 9797 respectivelyWhen 120572 = 095 the average service level obtained by thesampling approach is 9356 and those yielded by MIP-DF1 and MIP-DF2 are 9997 and 9781 respectively ForPoisson distribution concluding MIP-DF1 and MIP-DF2can improve the service level compared with the samplingapproach especially MIP-DF1 can obtain the highest servicelevel
Similarly Table 6 reports the computational results forinstances in which the demand at each emergency point isgenerated following Normal distribution When 120572 = 085the average service levels obtained by the sampling approachMIP-DF1 and MIP-DF2 are 8174 9925 and 9742respectively When 120572 = 09 the three values are 84579981 and 9743 respectively Moreover for 120572 = 095 thevalues are equal to 8732 9999 and 9794 respectivelyIt can be observed that the performance of the samplingapproach is the poorest in terms of the service level
By observing the above computational results on bench-marks and randomly generated instances we conclude that(1)The proposed approximated MIP formulations ie MIP-DF1 and MIP-DF2 are very time saving compared with thesampling approach (2) MIP-DF1 an MIP-DF2 can achievethe requirement of service level and improve the service levelby about 1547 and 1229 on average compared with thatof the sampling approach (3) MIP-DF1 is more conservativein terms of overhigh emergency service level close to 100and the system cost is also high which is about 568 largerthan that of MIP-DF2 (4) Compared with MIP-DF1 MIP-DF2 is less conservative and much more cost saving whileensuring an appropriately service level on average which isabout 97
In sum for medium and large problem instances (egShanghai) which are common in practice to guarantee therequired service level with less cost we recommendMIP-DF2as solution method However for very small instances with
8 Journal of Advanced Transportation
Table4Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mun
iform
distrib
ution
Set (|119868||119869|)
120572=085
120572=09
120572=095
Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
1(26)
27
897317
01
149
9906
01
107
9222
28
967373
01
155
9933
01
107
9167
27
106
7585
01
185
9987
01
107
9128
2(412
)47
169
7423
00
291
9919
00
207
9122
58
181
7438
01
303
9997
01
207
9363
50
199
7702
00
363
9982
00
207
9338
3(618
)72
258
7506
01
438
9976
01
312
9106
75279
8082
01
456
9996
01
312
9369
78304
8337
01
546
9997
01
312
9357
4(824)
99314
7956
01
553
9993
01
385
9206
111
338
8202
01
577
9999
01
385
9602
112
366
8492
01
697
9994
01
385
9620
5(1030)
132
415
8406
01
722
9896
01
512
9406
143
441
8322
02
752
9983
01
512
9692
142
476
8609
01
902
9993
01
512
9692
6(1236)
170
475
8249
02
843
9870
01
591
9367
181
504
8251
01
879
9982
01
591
9608
189
535
8726
01
1059
9997
01
591
9609
7(14
42)
215
573
8111
02
1006
9849
02
712
9606
243
609
8217
02
1048
9953
02
712
9589
233
644
8608
02
1258
9992
02
712
9582
8(1648)
248
686
7532
02
1185
9896
02
849
9725
284
727
8307
02
1233
9949
02
849
9613
281
768
9063
02
1473
9999
03
849
9622
9(1854)
313
691
8353
03
1236
9974
04
858
9709
351
731
8397
03
1290
9988
03
858
9765
352
760
8906
03
1560
9994
03
858
9758
10(2060)
372
794
8174
03
1409
9887
03
989
9657
428
843
8489
03
1469
9947
03
989
9773
432
892
9004
03
1769
9997
03
989
9738
11(2266
)451
830
8072
04
1493
9949
03
1031
9758
518
880
8328
04
1559
9982
03
1031
9723
516
930
8967
05
1889
9995
03
1031
9727
12(2472)
543
1025
7970
04
1771
9927
04
1267
9835
601
1083
8482
05
1843
9949
04
1267
9776
637
1146
9014
04
2203
9993
04
1267
9777
13(2678)
631
1026
8142
05
1819
9898
04
1274
9782
743
1087
8678
05
1897
9936
04
1274
9778
765
1146
9023
04
2287
9997
05
1273
9837
14(2884)
782
1101
7903
05
1949
9869
08
1361
9804
878
1165
8790
05
2033
9935
07
1361
9889
870
1227
9056
05
2453
9998
06
1361
9808
15(3090)
907
1091
8098
06
1997
9915
06
1367
9629
1018
1155
8329
06
2087
9981
06
1367
9807
1020
1225
8986
06
2537
9999
08
1367
9817
16(3296)
1062
1160
8798
07
2123
9944
07
1449
9787
1183
1229
8383
07
2219
9988
06
1449
9849
1193
1300
9105
06
2698
9998
06
1449
9815
17(34102)
1202
1298
7572
07
2328
9969
07
1614
9821
1377
1374
8860
08
2430
9997
07
1614
9824
1260
1452
9085
07
2940
9999
08
1614
9818
18(36108)
1394
1408
8121
08
2509
9947
08
1752
9809
1544
1492
8928
08
2617
9983
08
1752
9799
1548
1580
9173
08
3156
9999
08
1752
9809
19(38114
)1618
1508
8170
09
2676
9899
111877
9742
1760
1598
9130
09
2790
9934
08
1877
9908
1755
1688
9186
09
3360
9999
09
1877
9872
20(40120)
2206
1628
7719
102869
9989
102029
9873
2007
1732
8753
112990
9997
09
2029
9879
1904
1827
9138
103590
9998
09
2029
9875
21(42126)
2780
1622
8354
102900
9978
102017
9732
2609
1718
8854
103026
9876
102017
9818
2670
1812
9089
123655
9999
102017
9848
22(44132)
3189
1735
8494
113088
9985
152165
9857
2970
1840
8922
123220
9925
122165
9886
2873
1952
9206
113880
9999
112165
9850
23(46138)
3630
1937
8141
123372
9893
132406
9796
3222
2048
8937
123510
9934
142406
9906
3461
2170
8989
124199
10000
122406
9809
24(48144)
4226
1968
8226
133461
9932
132453
9814
3674
2085
8846
133605
9956
152453
9912
3556
2211
9300
134326
9999
132453
9819
25(50150)
4727
1919
8111
143456
9999
152407
9629
4019
2037
8699
143606
9999
142407
9857
4257
2161
9259
154356
9995
142407
9819
26(52156)
5300
1906
8096
153484
9878
172392
9658
4495
2021
8952
143640
9999
152392
9568
4486
2142
9182
1544
209999
152392
9854
27(54162)
5913
2096
8387
163751
9846
172617
9761
5117
2221
9045
163913
9998
172617
9544
4830
2355
9346
184723
9999
162617
9893
28(56168)
6310
2110
7758
173819
9933
192643
9829
5777
2234
9157
183987
9998
182643
9526
5550
2365
9045
184827
9999
172643
9831
29(58174)
6780
2175
7998
183937
9898
182720
9714
6303
2307
9022
20
4111
9969
192720
9623
6189
2444
9128
194981
10000
192720
9914
30(60180)
7593
2384
8321
194240
9967
20
2979
9841
6791
2527
8978
1944
199979
23
2979
9721
6878
2679
9143
21
5316
9999
20
2979
9869
31(62186)
8220
2281
7604
20
4157
9984
21
2856
9796
7734
2416
8934
21
4343
9997
21
2856
9679
7466
2557
9158
22
5273
9999
23
2856
9902
32(64192)
8766
2517
8378
21
4489
9932
25
3145
9729
8276
2671
8296
23
4680
9977
24
3145
9825
7897
2822
9089
24
5640
9999
22
3145
9891
33(66198)
10123
2532
7806
23
4555
9967
23
3169
9615
9010
2682
8947
24
4752
9989
24
3169
9703
9051
2835
9157
24
5742
9998
24
3169
9867
34(68204)
11010
2535
7934
24
4602
9999
29
3174
9631
9898
2685
8858
30
4806
9998
28
3174
9762
9863
2844
9236
26
5824
9999
25
3174
9889
35(70210)
11666
2714
8062
25
4862
9986
29
3391
9787
1066
62875
9137
26
5072
9999
29
3391
9716
10822
3043
9315
26
6122
9999
29
3391
9923
36(72216)
12011
2827
8104
26
5047
9991
28
3535
9798
1246
52984
9016
28
5263
9999
28
3535
9878
11667
3169
9253
30
6343
9999
29
3535
9896
37(74222)
11437
2826
8228
28
5096
9933
29
3543
9884
13582
3098
8895
29
5318
9999
32
3543
9888
12654
3178
9139
29
6427
10000
31
3543
9893
38(76228)
12435
2839
8152
29
5161
9897
29
3565
9788
13976
3103
8926
29
5388
9999
31
3565
9724
13837
3185
9255
33
6528
10000
31
3565
9895
39(78234)
1464
32945
8412
31
5331
9998
32
3695
9876
14794
3210
9005
31
5565
9999
34
3695
9713
14561
3313
9081
34
6735
10000
33
3695
9878
40(80240)
1604
22975
8510
32
5414
9957
36
3734
9796
16263
3346
8999
35
5654
9999
35
3734
9806
15869
3343
9248
35
6854
10000
35
3734
9873
Average
4482
1584
680
67
1228
397
9936
131978
796
82
4379
16913
8654
1329
626
9975
131978
79721
4295
1778
889
85
1335774
9997
131978
79775
Journal of Advanced Transportation 9
Table5Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mPo
isson
distr
ibution
set ( |119868 |
|119869 |)120572=
085120572=
09120572=
095Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
1(26)
27
907294
9967
9001
27
987770
9979
9013
24
103
9273
9999
9468
2(412
)46
170
7961
9921
9117
51
183
8305
9946
9128
49
201
9392
9998
9546
3(618
)68
258
7578
9938
8911
74280
8028
9968
9004
69
307
9182
9999
9658
4(824)
93315
8001
9928
9196
102
338
8438
9967
9172
92368
9468
9999
9613
5(1030)
120
413
8041
9939
9449
133
443
8447
9976
9272
121
479
9544
9999
9752
6(1236)
155
476
8084
9877
9554
167
508
8863
9991
9569
159
536
9456
9999
9734
7(14
42)
195
573
7876
9892
9663
213
609
9124
9998
9729
205
643
9314
10000
9768
8(1648)
232
687
7964
9997
9788
267
724
8523
9988
9687
238
772
9422
9998
9689
9(1854)
293
692
8267
9893
9809
327
730
8386
9999
9750
298
757
9228
9999
9755
10(2060)
356
796
7946
9899
9801
399
840
8299
9976
9686
360
885
9668
9999
9705
11(2266)
423
829
8201
9895
9756
489
877
8378
9987
9788
448
927
9476
9999
9812
12(2472)
512
1023
8338
9913
9853
554
1085
8501
9999
9881
514
1148
9398
9999
9888
13(2678)
592
1045
8416
9907
9873
687
1088
7998
9996
9802
607
1152
9504
9999
9821
14(2884)
723
1099
7946
9915
9788
838
1166
8056
9992
9829
776
1230
9546
9999
9839
15(3090)
855
1089
7915
9964
9703
942
1156
8432
9984
9778
903
1225
9223
9998
9785
16(3296)
1009
1158
8024
9897
9769
1163
1231
8500
9996
9824
1045
1222
9199
9999
9835
17(34102)
1165
1299
8034
9919
9779
1285
1375
8495
9937
9808
1190
1457
9258
10000
9846
18(36108)
1254
1411
8176
9943
9885
1479
1492
8561
9968
9868
1345
1579
9286
9999
9875
19(38114
)1504
1506
7970
9923
9842
1651
1594
8167
9972
9846
1564
1694
9376
9999
9786
20(40120)
1706
1635
7819
9916
9783
1854
1734
8607
9968
9776
1778
1825
9285
9904
9718
21(42126)
2054
1622
8354
9918
9834
2357
1720
8297
9978
9833
2226
1856
9501
9999
9834
22(44132)
2400
1736
8494
9922
9846
2852
1844
8269
9999
9816
2680
1946
9486
9999
9820
23(46138)
2651
1934
8141
9953
9793
3159
2049
8492
9969
9799
3072
2172
9358
10000
9813
24(48144)
3004
1961
8226
9947
9681
3522
2086
8513
9998
9706
3102
2209
9488
9999
9724
25(50150)
3308
1924
8111
9965
9746
3898
2038
8264
9999
9788
3639
2161
9546
9999
9813
26(52156)
3669
1911
8096
9939
9855
4305
2027
8722
9967
9874
4453
2142
9185
9999
9876
27(54162)
4418
2095
8387
9927
9754
4771
2219
8341
9989
9749
4806
2353
9289
9999
9755
28(56168)
5367
2113
8458
9968
9848
5356
2237
8452
9978
9855
5221
2369
9228
10000
9868
29(58174)
6217
2186
7998
9989
9694
5942
2315
8316
9996
9699
5853
2371
9167
9999
9705
30(60180)
7263
2382
8321
9991
9837
6527
2543
8359
9998
9886
6214
2678
9369
9997
9886
31(62186)
8214
2346
8404
9936
9888
7813
2430
8321
9999
9846
6876
2560
9358
9999
9847
32(64192)
8863
2518
8378
9938
9807
8106
2677
8356
9999
9833
7877
2822
9408
10000
9835
33(66198)
9985
2532
8406
9917
9709
9684
2689
8497
9996
9746
8088
2838
9507
9999
9749
34(68204)
11264
2534
8543
9908
9897
10986
2685
8809
9996
9782
8469
2843
8844
9998
9780
35(70210)
12035
2715
8258
9937
9799
12098
2886
8569
9991
9809
9946
3041
9459
9999
9813
36(72216)
12803
2825
8104
9929
9827
13076
3010
8651
9994
9834
11013
3170
9468
9999
9865
37(74222)
13021
2830
8228
9949
9826
13261
3009
8376
9999
9826
12014
3172
9346
9999
9888
38(76228)
13862
2843
8152
9933
9807
13995
3028
8760
9998
9841
12133
3187
9706
9999
9876
39(78234)
15387
2953
8312
9964
9799
14827
3133
9001
9995
9783
13381
3304
8913
9999
9799
40(80240)
16974
2975
8510
9985
9767
15256
3249
8217
9993
9781
15867
3347
9111
9999
9791
Average
4352
15875
8143
9934
9708
4362
16856
8437
9985
9707
3968
17763
9356
9997
9781
10 Journal of Advanced Transportation
Table6Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mno
rmaldistr
ibution
set ( |119868 |
|119869 |)120572=
085120572=
09120572=
095Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
1(26)
27
907294
9964
9117
27
947844
9970
9202
2558
106
8422
10000
9208
2(412
)46
170
7961
9921
9283
48
183
8328
9982
9283
4508
183
8823
10000
9370
3(618
)68
258
7578
9916
9228
68
280
7798
9971
9417
6803
307
7658
9998
9447
4(824)
93315
8000
9899
9367
96337
8379
9979
9467
9373
368
8438
9999
9579
5(1030)
120
413
7841
9927
9437
124
443
8763
9963
9476
1262
499
8850
9998
9708
6(1236)
155
476
7921
9919
9519
162
504
8124
9982
9582
16021
536
8954
10000
9756
7(14
42)
195
573
7976
9933
9598
196
609
8287
9975
9619
2011
646
8467
9999
9767
8(1648)
232
687
8031
9918
9705
235
727
8115
9965
9650
2404
772
8278
9999
9765
9(1854)
293
692
8086
9898
9759
291
734
8464
9974
9755
29815
796
8617
9998
9835
10(2060)
356
796
7984
9937
9770
358
842
8223
9988
9765
37973
921
8846
10000
9838
11(2266
)421
829
8101
9940
9755
436
877
8211
9986
9786
4514
8959
8497
9999
9789
12(2472)
514
1023
8033
9951
9785
501
1085
8162
9983
9788
52073
1147
8321
9999
9824
13(2678)
592
1023
8138
9916
9885
599
1087
8411
9970
9845
62937
1194
8609
9998
9795
14(2884)
728
1099
8235
9913
9915
722
1168
8435
9987
9806
75053
1270
8567
9997
9834
15(3090)
855
1089
8187
9922
9861
880
1157
8376
9976
9796
89222
1300
8611
9996
9798
16(3296)
1009
1158
8770
9897
9788
966
1231
8845
9984
9823
106288
1359
8997
9999
9848
17(34102)
1154
1299
8434
9925
9849
1137
1377
8601
9980
9798
123869
1493
8798
9996
9877
18(36108)
1407
1411
8321
9927
9866
1339
1495
8301
9965
9822
133807
1601
8567
9999
9834
19(38114
)1702
1506
8206
9915
9805
1672
1598
8977
9978
9845
152999
1694
8998
9999
9847
20(40120)
2091
1635
8528
9918
9889
1916
1737
8805
9976
9837
1890
021833
9012
9999
9791
21(42126)
2843
1622
8550
9934
9913
2432
1718
8886
9974
9813
2399
131815
9008
10000
9815
22(44132)
3300
1736
8449
9940
9859
2795
1843
8786
9977
9756
253106
1948
8997
9999
9826
23(46138)
3741
1934
8133
9911
9876
3143
2050
8876
9981
9785
274583
2173
9014
9999
9858
24(48144)
4531
1961
8245
9909
9846
3515
2090
8344
9993
9819
308089
2211
8661
9999
9883
25(50150)
4715
1924
8111
9929
9888
3959
2041
8279
9996
9837
3510
932158
8527
9999
9827
26(52156)
5223
1911
8096
9898
9857
4301
2025
8251
9997
9757
421747
2146
8497
9999
9914
27(54162)
5994
2095
8387
9908
9775
4835
2226
8497
9974
9780
4614
432359
8695
9999
9837
28(56168)
6406
2111
8411
9906
9798
5336
2240
8558
9991
9820
494113
2371
9249
9999
9816
29(58174)
6897
2176
8012
9916
9791
5948
2311
8211
9984
9765
624835
2445
8531
10000
9851
30(60180)
7436
2391
8111
9913
9766
6558
2533
8305
9968
9795
696116
2682
8575
9999
9799
31(62186)
8118
2282
8119
9928
9806
7069
2422
8417
9973
9873
797768
2564
8667
9999
9846
32(64192)
8895
2519
8223
9919
9757
7857
2675
8631
9982
9769
864813
2827
8976
10000
9905
33(66198)
11235
2533
8563
9928
9811
9162
2688
8547
9979
9828
973089
2843
8657
9999
9846
34(68204)
11587
2534
8249
9926
9847
9549
2686
8798
9938
9856
10523
3041
8999
10000
9877
35(70210)
1164
92715
8098
9895
9872
10345
2880
8602
9998
9869
11272
3052
9021
9999
9897
36(72216)
12224
2825
8403
9958
9770
11165
3002
8657
9999
9839
12456
3175
9113
9999
9864
37(74222)
12978
2830
8357
9937
9795
11850
3005
8496
9998
9894
13458
3176
8827
9999
9905
38(76228)
13426
2844
8267
9966
9759
12795
3019
8604
9999
9797
15155
3121
8978
9999
9853
39(78234)
15003
2955
8303
9956
9867
13816
3132
8588
9996
9885
15041
3310
9016
9999
9912
40(80240)
16495
2978
8257
9954
9826
14896
3160
8495
9997
9837
16563
3349
8957
9999
9933
Average
4619
15855
8174
9925
9742
4077
16828
8457
9981
9743
4350
17938
8732
9999
9794
Journal of Advanced Transportation 11
high service level requirement we recommend MIP-DF1 forsolutions
6 Conclusion
This paper investigates an ambulance location problem withambiguity set of demand in which the probability distribu-tion of the uncertain demand is unknown In such a data-driven environment only the mean and covariance demandsat emergency points are knownThe problem is to determinethe base locations and the number of ambulances at eachbase aiming at minimizing the total cost associated withlocating bases and assigning ambulances We propose adistribution-free model with chance constraints Then twoapproximated MIP formulations with different approxima-tion methods of chance constraints are proposed which arebased on different ambiguity sets of the unknown probabilitydistribution Numerical experiments on benchmarks and120 randomly generated instances are conducted and thecomputational results show that the first approximated MIPformulation is much conservative with overhigh system costwhile the second approximatedMIP formulation ismore costsaving with service level appropriately ensured
In future research wemay consider the following relevantissues First the uncertainty of driving time from a base toan emergency point shall be analysed Besides unexpected orsudden disasters may be taken into consideration Moreoverother uncertainties should be considered such as travel timethe number of available vehicles and so on The stochasticdependence should also further considered
Data Availability
The data of the benchmark instances in our manuscript canbe obtained in Nickel et al [8] For the test instances thedata is randomly generated and the way for generating data isstated in our manuscript The datasets generated during thecurrent study are available from the corresponding author onreasonable request
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This work was supported by the National Natural Sci-ence Foundation of China (NSFC) under Grants 7142800271531011 71571134 and 71771048 This work was also sup-ported by the Fundamental Research Funds for the CentralUniversities
References
[1] L Brotcorne G Laporte and F Semet ldquoAmbulance loca-tion and relocation modelsrdquo European Journal of OperationalResearch vol 147 no 3 pp 451ndash463 2003
[2] X Li Z Zhao X Zhu and T Wyatt ldquoCovering modelsand optimization techniques for emergency response facilitylocation and planning a reviewrdquo Mathematical Methods ofOperations Research vol 74 no 3 pp 281ndash310 2011
[3] C Toregas R Swain C ReVelle and L Bergman ldquoThe locationof emergency service facilitiesrdquoOperations Research vol 19 no6 pp 1363ndash1373 1971
[4] R Church and C ReVelle ldquoThe maximal covering locationproblemrdquo Papers in Regional Science vol 32 no 1 pp 101ndash1181974
[5] M Gendreau G Laporte and F Semet ldquoSolving an ambulancelocation model by tabu searchrdquo Location Science vol 5 no 2pp 75ndash88 1997
[6] A Ingolfsson S Budge and E Erkut ldquoOptimal ambulancelocation with random delays and travel timesrdquo Health CareManagement Science vol 11 no 3 pp 262ndash274 2008
[7] M Gendreau G Laporte and F Semet ldquoThemaximal expectedcoverage relocation problem for emergency vehiclesrdquo Journal ofthe Operational Research Society vol 57 no 1 pp 22ndash28 2006
[8] S Nickel M Reuter-Oppermann and F Saldanha-da-GamaldquoAmbulance location under stochastic demand A samplingapproachrdquo Operations Research for Health Care vol 8 pp 24ndash32 2016
[9] P Beraldi M E Bruni and D Conforti ldquoDesigning robustemergency medical service via stochastic programmingrdquo Euro-pean Journal of Operational Research vol 158 no 1 pp 183ndash1932004
[10] N Noyan ldquoAlternate risk measures for emergency medicalservice system designrdquo Annals of Operations Research vol 181pp 559ndash589 2010
[11] M R Wagner ldquoStochastic 0-1 linear programming underlimited distributional informationrdquoOperations Research Lettersvol 36 no 2 pp 150ndash156 2008
[12] EDelage andYYe ldquoDistributionally robust optimization undermoment uncertainty with application to data-driven problemsrdquoOperations Research vol 58 no 3 pp 595ndash612 2010
[13] J T Essen J L Hurink S Nickel and M Reuter ldquoModels forambulance planning on the strategic and the tactical levelrdquo BetaResearch School for Operations Management and Logistics p 272013
[14] E Erkut A Ingolfsson and G Erdogan ldquoAmbulance locationfor maximum survivalrdquoNaval Research Logistics (NRL) vol 55no 1 pp 42ndash58 2008
[15] S C Chapman and J A White ldquoProbabilistic formulations ofemergency service facilities location problemsrdquo in Proceedingsof ORSATIMS Conference San Juan Puerto Rico 1974
[16] A Ben-Tal D Den Hertog A De Waegenaere B Melenbergand G Rennen ldquoRobust solutions of optimization problemsaffected by uncertain probabilitiesrdquo Management Science vol59 no 2 pp 341ndash357 2013
[17] MNg ldquoDistribution-free vessel deployment for liner shippingrdquoEuropean Journal of Operational Research vol 238 no 3 pp858ndash862 2014
[18] M Ng ldquoContainer vessel fleet deployment for liner shippingwith stochastic dependencies in shipping demandrdquo Transporta-tion Research Part B Methodological vol 74 pp 79ndash87 2015
[19] R Jiang and Y Guan ldquoData-driven chance constrained stochas-tic programrdquo Mathematical Programming vol 158 no 1-2 SerA pp 291ndash327 2016
[20] C-M Lee and S-L Hsu ldquoThe effect of advertising on thedistribution-free newsboy problemrdquo International Journal ofProduction Economics vol 129 no 1 pp 217ndash224 2011
[21] K Kwon and T Cheong ldquoA minimax distribution-free proce-dure for a newsvendor problem with free shippingrdquo EuropeanJournal of Operational Research vol 232 no 1 pp 234ndash2402014
12 Journal of Advanced Transportation
[22] Y Zhang S Shen and S A Erdogan ldquoDistributionally robustappointment scheduling with moment-based ambiguity setrdquoOperations Research Letters vol 45 no 2 pp 139ndash144 2017
[23] Y Zhang S Shen and S A Erdogan ldquoSolving 0-1semidefinite programs for distributionally robust allocationof surgery blocksrdquo Social Science Electronic Publishing 2017httpsssrncomabstract=2997524
[24] F Zheng J He F Chu and M Liu ldquoA new distribution-freemodel for disassembly line balancing problem with stochastictask processing timesrdquo International Journal of ProductionResearch pp 1ndash13 2018
[25] J Cheng E Delage and A Lisser ldquoDistributionally robuststochastic knapsack problemrdquo SIAM Journal on Optimizationvol 24 no 3 pp 1485ndash1506 2014
[26] M Liu D Yang and F Hao ldquoOptimization for the locationsof ambulances under two-stage life rescue in the emergencymedical service a case study in Shanghai ChinardquoMathematicalProblems in Engineering vol 2017 Article ID 1830480 14 pages2017
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
2 Journal of Advanced Transportation
Moreover how to cover as many emergencies as possibleto guarantee a high patient service level (ie the portion ofthe satisfied emergency points among all emergency points)has always been a main goal of EMS operators Motivatedby the above observation this work focuses on a stochasticambulance location problem with only partial informationof demand points ie the mean and covariance matrix ofthe demands Besides we introduce a new individual chanceconstraint which implies a minimum probability that eachemergency demand has to be satisfied For the problema new distribution-free model is presented and then twoapproximated formulations are proposed The contributionof this work mainly includes the following
(1) For the studied stochastic ambulance location prob-lem we introduce a new individual chance con-straint (ie safety level) guaranteeing each emergencydemand satisfied with a least probability while Nickelet al [8] use a coverage constraint where some serviceof individual emergencies may suffer insufficiency
(2) A new distribution-free model based on individualchance constraints is proposed To our best knowl-edge it is the first distribution-freemodel for stochas-tic ambulance location problem
(3) To solve the distribution-free model two approx-imated mixed integer programming (MIP) formu-lations are developed Experimental results showthat our approximated formulations are efficient andeffective for large size instances compared to thatproposed by Nickel et al [8] in terms of both com-putational time and service level
The remainder of this paper is organized as followsSection 2 gives a brief literature review In Section 3 we givethe problem description and propose a new distribution-freemodel In Section 4 two approximated MIP formulationsare proposed Computational results on benchmarks and120 randomly generated instances are reported in Section 5Section 6 summarizes this work and states future researchdirections
2 Literature Review
The deterministic ambulance location problem has been wellstudied in literature (eg [2 5 13] etc) Besides there havebeen some works addressing ambulance location problemunder uncertainty (eg [1 6 14] etc) Since our study fallswithin the scope of stochastic ambulance location problem inthe following subsections we first review existing studies onambulance location problem with uncertain demand Thenwe review the literature studying general and specific stochas-tic optimization problems with distribution-free approaches
21 Ambulance Location Problem with Uncertain DemandAmbulance location problem under uncertain demand hasbeen investigated by many researchers Most existing worksaddress the uncertainty with given scenarios or knownprobability distribution
Chapman and White [15] first investigate the ambulancelocation problem with uncertain demand in which the
complete information on probability distributions is assumedto be known Beraldi et al [9] study the emergency medicalservice location problemwith uncertain demand tominimizethe total cost with amarginal probability distribution Beraldiand Bruni (2008) investigate the emergency medical servicefacilities location problem with stochastic demand based ongiven set of scenarios and known probability distributionA stochastic programming formulation with probabilisticconstraints is proposed Noyan [10] studies the ambulancelocation problem with uncertain demand on given scenariosand probability distribution Then two stochastic optimiza-tion models and a heuristic are proposed
Recently Nickel et al [8] first study the joint optimizationof ambulance base location and ambulance employmentTheobjective is to minimize the total constructing base costsand ambulance employment costs It is assumed that variousscenarios are given beforehand A coverage constraint isintroduced in the paper They propose a sampling approachthat solves a finite number of scenario samples to obtaina feasible solution of the original problem But with thecoverage constraint some emergency demands risk insuffi-cient individual service or cannot be served For the studywe propose a new distribution-free model for stochasticambulance location problem
22 Distribution-Free Approaches In data-driven settingsthe probability distributions of uncertain parameters maynot always be perfectly estimated [16] Therefore in the lastdecade there have beenmany solution approaches developedto address stochastic problems under partial distributionalknowledgeMost related researches focus on the distribution-free approach via considering chance constraintsWagner [11]studies a stochastic 0-1 linear programming under partialdistribution information ie min119883isin01119899c1015840x a1015840119895x le119887119895 119895 = 1 119898 where a119895 are random vectors withunknown distributions The only information on a119895 aretheir moments up to order 119896 A robust formulation asa function of 119896 is given Given the known second-ordermoment knowledge ie 119896 = 2 an approximated formulationis developed as min119883isin01119899c1015840x radicx1015840Γ119895x le radic119901119895(1 minus 119901119895)(119887119895 =E[a119895]1015840x) 119895 = 1 119898 where Γ119895
119894119896= E[(119886119894119895 minus E[119886119894119895])(119886119896119895 minus
E[119886119896119895])] forall119894 119896 = 1 119899 Delage and Ye [12] investigate thestochastic program with limited distribution informationand they propose a new moment-based ambiguity set whichis assumed to include the true probability distribution todescribe the uncertainties There have been various workssuccessfully applying the distribution-free approaches Ng[17] investigates a stochastic vessel deployment problem forliner shipping in which only the mean standard deviationand an upper bound of demand are known A distribution-free optimization formulation is proposed Based on that Ng[18] studies a stochastic vessel deployment problem for theliner shipping industry where only the mean and varianceof the uncertain demands are known New models areproposed and the provided bounds are shown to be sharpunder uncertain environment The stochastic dependenciesbetween the shipping demands are considered
Jiang and Guan [19] develop approaches to solve stochas-tic programs with data-driven chance constraints Two types
Journal of Advanced Transportation 3
of confidence sets for the possible probability distributionsare proposed For more distribution-free formulation appli-cations please see Lee and Hsu [20] Kwon and Cheong [21]etc for the stochastic inventory problem and Zhang et al[23] for stochastic allocating surgeries problem in operatingrooms and Zheng et al [24] for stochastic disassembly linebalancing problem To the best of our knowledge there is noresearch for the stochastic ambulance location problem withonly partial information on the uncertain demand
3 Problem Description and Formulation
In this section we first describe the considered problem andthen propose a new distribution-free model
31 Problem Description There is a given set of candidatebase locations 119868 = 1 2 |119868| and a set of potentialemergency demand points 119869 = 1 2 |119869| The emergencypoints may refer to a road a part of the urban area and avillage Base 119894 isin 119868 is said to be covering an emergency point119895 if the driving time 119905119894119895 between 119894 and 119895 is no more than apredetermined value119879The set 119868119895 of candidate bases coveringemergency point 119895 is denoted as 119868119895 = 119894 isin 119868 | 119905119894119895 le 119879
The number of ambulances required by an emergencypoint 119895 isin 119869 is denoted as 119889119895 which depends on the severityof the practical situationWe consider uncertain emergencieswhich are estimated or predicted by partial distributionalinformation thus the demand is regarded to be ambiguousBesides we focus on the case where the historical data cannotbe represented by a precise probability distribution It isassumed that the customer demands are independent of eachother
The objective of the problem is to select a subset ofbase locations and determine the number of ambulances ateach constructed base in order to minimize the total baseconstruction cost and ambulance costThroughout this paperwe assume that only the mean and covariance matrix ofdemands are known Moreover a predefined safety level 120572119895is given for each potential emergency point 119895 isin 119869 That is thenumber of ambulances serving each emergency point 119895 isin 119869 islarger than or equal to its demand with a least probability of12057211989532 Chance Constraint Construction In this section weintroduce a chance constraint to guarantee the safety levelof each emergency demand point In the following 119910119894119895 is adecision variable denoting the number of ambulances servingpoint 119895 isin 119869 from base 119894 isin 119868119895 and 119889119895 is the uncertain demandat point 119895 isin 119869 The chance constrained inequality is presentedas follows
ProbP (sum119894isin119868119895
119910119894119895 ge 119889119895) ge 120572119895 forall119895 isin 119869 (1)
where ProbP(sdot) denotes the probability of the event inparentheses under any potential probability distribution PConstraint (1) ensures that the number of ambulances servingemergency point 119895 isin 119869 is no less than its demand with a leastprobability of 120572119895 (ie safety level)
33 Distribution-Free Formulation In the following we givebasic notations define decision variables and propose thedistribution-free formulationDF for the ambulance locationproblem with uncertain demand
Parameters
119894 index of base locations119895 index of emergency points119868 set of candidate base locations119869 set of possible emergency demand points119905119894119895 driving time between base 119894 isin 119868 and demand point119895 isin 119869119879 maximum driving time for serving any emergencycall119868119895 set of candidate bases that can cover emergencypoint 119895 isin 119869 ie 119868119895 = 119894 isin 119868 | 119905119894119895 le 119879119869119894 set of potential emergency points covered by base119894 isin 119868 ie 119869119894 = 119895 isin 119869 | 119905119894119895 le 119879119891119894 fixed construction cost for installing a base atlocation 119894 isin 119868119892119894 fixed cost associated with an ambulance to belocated at 119894 isin 119868119889119895 number of (stochastic) ambulances requested byemergency point 119895 isin 119869M a sufficiently large positive number
Decision Variables
119909119894 a binary variable equal to 1 if a base is constructedat 119894 isin 119868 0 otherwise119911119894 number of ambulances assigned to possible baselocation 119894 isin 119868119910119894119895 number of ambulances serving emergency point119895 isin 119869 from possible base location 119894 isin 119868119895
Distribution-Free Model [DF]
[DF]
min sum119894isin119868
(119891119894 sdot 119909119894 + 119892119894 sdot 119911119894)st Constraint (1)
(2)
119911119894 le M sdot 119909119894 forall119894 isin 119868 (3)
sum119895isin119869119894
119910119894119895 le 119911119894 forall119894 isin 119868 (4)
119909119894 isin 0 1 forall119894 isin 119868 (5)
119911119894 isin Z+ forall119894 isin 119868 (6)
119910119894119895 isin Z+ forall119894 isin 119868 119895 isin 119869119894 (7)
4 Journal of Advanced Transportation
The objective function denotes the goal to minimize thetotal cost consisting of two parts (i) the cost for constructingbases ie sum119894isin119868 119891119894 sdot 119909119894 and (ii) the cost for deploying ambu-lances ie sum119894isin119868 119892119894 sdot 119911119894
Constraint (3) ensures that ambulances can only belocated at the opened bases Constraint (4) ensures that thenumber of ambulances sent to serve emergency points frombase 119894 isin 119868 does not exceed the total number of ambulanceslocated at 119894 isin 119868 Constraints (5)-(7) are the restrictions ondecision variables
4 Solution Approaches
The proposed distribution-free model is difficult to solvewith the commercial software due to chance constraints Inthis section we propose two approximatedMIP formulationsbased on those in Wagner [11] and in Delage and Ye [12]respectively In the following subsections we present the twoapproximated MIP formulations
41 Approximated MIP Formulation MIP-DF1 In this partwe first describe a widely used ambiguity set to describe theuncertainty Then an approximated MIP formulation MIP-DF1 is proposed
Given a set of independent historical data samples 119889119904|119878|119904=1of random vectors of demands where 119878 is the set of sampleindexes the mean vector 120583 and the covariance matrix Σ ofdemands can be estimated as follows
120583 = 1|119878|sum119904isin119878119889119904Σ = 1|119878|sum119904isin119878 (119889119904 minus 120583) (119889119904 minus 120583)⊤
(8)
where (sdot)⊤ implies the transposition of the vector in paren-theses Then an ambiguity set of all probability distributionsof demandsP1(120583 Σ) can be described as the following
P1 (120583 Σ) = P EP [119889] = 120583EP [(119889 minus 120583) (119889 minus 120583)⊤] = Σ (9)
whereP denotes a possible probability distribution satisfyingthe given conditions and EP[sdot] denotes the expected valueof the number in parentheses Then the chance constraint (1)with the ambiguity setP1 can be presented as follows
ProbP (sum119894isin119868119895
119910119894119895 ge 119889119895) ge 120572119895 forall119895 isin 119869 P isin P1 (10)
According to Wagner (2010) and Ng [18] constraint (1) canbe conservatively approximated by the following
sum119894isin119868119895
119910119894119895 ge 120583119895 + 120590119895 sdot radic 120572119895(1 minus 120572119895) forall119895 isin 119869 (11)
where 120590119895 = radicΣ119895119895 and Σ119895119895 denotes the 119895-th element in the 119895-thcolumn of matrix Σ In terms of conservative approximation
all solutions satisfying constraint (11) must satisfy constraint(1) Then we describe the approximated MIP formulationMIP-DF1 toDF combined with constraint (11)
[MIP-DF1]
min sum119894isin119868
(119891119894 sdot 119909119894 + 119892119894 sdot 119911119894)st Constraints (4) - (8) (12)
(12)
MIP-DF1 can be exactly solved by the commercialsoftware such as CPLEX As we can observe from thecomputational results in Section 5 it can obtain a relativelyhigh service level However the system cost is also high Inorder to save the system cost another approximated MIPformulation of DF is then proposed in the next subsection
42 Approximated MIP Formulation MIP-DF2 AmbiguitysetP1 focuses on exactly matching the mean and covariancematrix of uncertain parameters However in practice theremay be considerable estimation errors in the mean andcovariance matrix To take the inevitable estimation errorsinto consideration Delage and Ye [12] introduce a newmoment-based ambiguity set Therefore we in the followingemploy the moment-based ambiguity set
P2 (120583 Σ 1205741 1205742)= P (EP [119889] minus 120583)⊤ Σminus1 (EP [119889] minus 120583) le 1205741
EP [(119889 minus 120583) (119889 minus 120583)⊤] ⪯ 1205742Σ (13)
where 1205741 ge 0 and 1205742 ge 0 are two parameters of ambiguityset P2(120583 Σ 1205741 1205742) with the following assumptions (i) thetrue mean vector of demands is within an ellipsoid of sizeproportion to 1205741 centered at 120583 and (ii) the true covariancematrix of demands is in a positive semidefinite cone boundedby amatrix inequality of 1205742ΣThe ambiguity set describes howlikely the uncertain parameters are to be close to the meanin terms of the correlation [12] Besides under the moment-based ambiguity set various stochastic programs with partialdistributional information have been successfully modeledand solved [22 23 25]
According to the approximation method in Zhang et al[23] chance constraint (1) can be approximated by
radic 11 minus 119886 minus 119887 sdot (1 + radic1 minus 120572119895120572119895 sdot 119887) sdot 120590119895le radic1 minus 120572119895120572119895 sdot (sum
119894isin119868119895
119910119894119895 minus 120583119895) 119895 isin 119869(14)
where
119886 = 1 minus 1205741 + 11205742 minus 1205741 119887 = 12057411205742 minus 1205741
(15)
Journal of Advanced Transportation 5
Table 1 120572 = 095|119868| = |119869| Sampling approach MIP-DF1 MIP-DF2
CT Obj Sel () CT Obj Sel () CT Obj Sel ()1 7 49 23 9626 01 36 10000 01 23 100002 7 51 22 9863 01 35 10000 01 24 99823 6 41 14 9692 01 30 10000 01 14 98504 6 43 13 9713 02 29 10000 01 12 97425 6 57 12 7996 01 30 10000 02 12 100006 8 58 21 9850 01 24 10000 02 27 10000
Average 50 175 9457 01 307 10000 01 187 9929
The second approximated MIP formulationMIP-DF2 toDF is presented as follows
[MIP-DF2]
min sum119894isin119868
(119891119894 sdot 119909119894 + 119892119894 sdot 119911119894)st Constraints (4) - (8) (15)
(16)
Notice that when 1205741 = 0 and 1205742 = 1 inequality (14)reduces to (11) in MIP-DF1 which implies that MIP-DF1 isa special case of MIP-DF2
5 Computational Experiments
In this section the performance of our proposed formula-tions first evaluated and compared to the sampling approachproposed in Nickel et al [8] Specifically given a requiredservice level 120572 we apply our MIP formulations by restrictingthe chance constraint for each emergency point to satisfya safety level 120572119895 = 120572 while we adopt the samplingapproach by restricting the coverage constraint with 120572 Thencomputational results on 120 randomly generated instancesbased on the information on emergencies in Shanghai arereported Our approximated MIP formulations and thesampling approach are solved by coding in MATLAB 2014bcalling CPLEX 126 solver All numerical experiments areconducted on a personal computer with Core I5 and 330GHzprocessor and 8GB RAMunder windows 7 operating systemThe computational time for sampling approach is limited to3600 seconds
51 Out-of-Sample Test FollowingZhang et alrsquos [22]method-ology we examine the out-of-sample performance of solu-tions obtained by our proposed approximated MIP formula-tions and the sampling approach [8] The out-of-sample testincludes three steps
(1) The base locations and the employment of ambu-lances at each base are determined with known meanvector and covariance matrix of demands which isnamed as in-sample test
(2) 1000 scenarios are sampled from the underlying truedistribution which represent realizing the uncertaindemand
(3) Based on the solution obtained in step (1) ambu-lances are assigned to serve emergency points inthe 1000 scenarios to maximize the (out-of-sample)service level
The (out-of-sample) service level is defined as the ratioof the number of emergency points with demand satisfiedto the total number of emergency points in 1000 scenariosIn the following subsections the benchmarks are tested andthen three common distributions are employed ie uniformdistribution Poisson distribution and normal distributionrespectively
52 Benchmark Instances We first test the benchmarks inNickel et al [8] to compare our proposed two approximatedMIP formulationswith the sampling approach InNickel et al[8] they assume that demand nodes are also potential baselocations and the number of ambulances required by eachemergency point is chosen from set 0 1 2with given proba-bility distribution Computational results on benchmarks arereported in Tables 1 and 2
Table 1 reports the computational results with servicelevel requirement120572 = 095 In Table 1 CTObj and Sel denotethe computational time objective value (ie the system cost)and the service level respectively The first column includesthe indexes of instances and the second indicates the numberof nodes It shows that the average service level obtainedby the sampling approach is 9457 which is smaller thanthe required service level ie 85 Specifically the servicelevel of instance 5 obtained by the sampling approach is only7996 However MIP-DF1 and MIP-DF2 can guarantee theservice levels for all instances higher than the requirementConcluding from Table 1 we can observe that (i) the averagecomputational time of sampling approach is about 50 timesas those of the two approximated MIP formulations and (ii)MIP-DF1 and MIP-DF2 improve the average service level byabout 574 and 499 compared with sampling approachhowever (iii) system costs obtained by MIP-DF1 and MIP-DF2 are respectively 7543 and 686 higher than that ofsampling approach
Table 2 reports the computational results with servicelevel requirement 120572 = 099 For instances 2-6 the ser-vice levels obtained by sampling approach cannot meet therequirement ie 99 Besides it shows that the averagecomputational time of sampling approach is 48 which isabout 48 times greater than those of MIP-DF1 andMIP-DF2
6 Journal of Advanced Transportation
Table 2 120572 = 099|119868| = |119869| Sampling approach MIP-DF1 MIP-DF2
CT Obj Sel () CT Obj Sel () CT Obj Sel ()1 7 49 24 9979 01 64 10000 01 25 100002 7 48 23 9856 01 63 10000 01 24 100003 6 51 15 9717 01 55 10000 01 16 100004 6 47 15 9275 01 52 10000 01 13 100005 6 40 14 9000 01 55 10000 01 14 100006 8 52 23 9785 01 45 10000 01 27 10000
Average 48 190 9343 01 557 10000 01 198 10000
Table 3 Solutions to instance 4 in Nickel et al [8] for the given scenario
Sampling approach MIP-DF1 MIP-DF2Emergency points 1 2 3 4 5 6 Emergency points 1 2 3 4 5 6 Emergency points 1 2 3 4 5 6
Demand 1 0 2 0 2 0 Demand 1 0 2 0 2 0 Demand 1 0 2 0 2 0Base locations Ambulance employment Ambulance employment Ambulance employment1 0 0 02 1 1 0 3 1 23 1 1 26 1 2 2 04 0 0 05 0 6 06 2 2 0 4 2
The average service level is 9343 Although the averageservice levels obtained by MIP-DF1 and MIP-DF2 are both100 the average system cost obtained by MIP-DF1 is about3 times larger than that of sampling approach Besides theaverage system cost obtained by MIP-DF2 is 421 higherthan that of sampling approach
Concluding for the benchmarks described in Nickel etal [8] we can observe that (1) MIP-DF1 and MIP-DF2 canbe solved in far less computational time compared with thesampling approach (2) the sampling approach cannot meetthe requirement of service level for all benchmarks whilethe two MIP formulations can (3) the average system cost ofMIP-DF2 is very close to that of sampling approach (4) withthe increase of 120572 the average system costs obtained by MIP-DF2 is getting closer to that obtained by sampling approach
From Tables 1 and 2 we can observe that in spite ofthe lower system cost obtained by the sampling approachthere may exist extreme situations that demands of someemergency points are not satisfied Take one benchmark inNickel et al [8] ie instance 4 where there are 6 nodesA scenario in which demand at each emergency point isgenerated from the given probability distribution is shown inFigure 1 We can observe from Figure 1 that under the givenscenario the demand of each possible emergency point is1 0 2 0 2 0The solutions obtained by sampling approachMIP-DF1 andMIP-DF2 are shown in Table 3 which includesbase locations and ambulance employment obtained by in-sample test and ambulance assignment under given scenario
In Table 3 the first column includes the indexes of poten-tial base locations The second column represents the baselocations and the ambulance employment at each location
>1 = 1
>2 = 0
>3 = 2
>4 = 0
>5 = 2
>6 = 0
1
2 4
3 5
6
Figure 1 Instance 4 in Nickel et al [8] under given scenario
in which a location with 0 ambulance implies that it is notselected for base construction From Table 3 we can observethat the in-sample solutions obtained by sampling approachis to construct bases at locations 2 3 and 6 where the numberof ambulances is 1 1 and 2 respectively Besides undergiven scenario there is one ambulance serving emergencypoint 1 from base location 3 and one ambulance servingemergency point from base location 2 and 2 ambulancesserving emergency point 5 from base location 6 It shows thatthe requirement of ambulances at emergency point 3 is notfully satisfied The insufficient service may cause serious lossof life However we can find that the solutions obtained byMIP-DF1 andMIP-DF2 can provide sufficient ambulances forall emergency points
In sum to guarantee the required service level withless cost MIP-DF2 is recommended for the solutions with
Journal of Advanced Transportation 7
Nickel et al [8]rsquos data set Moreover in the benchmarks itis assumed that all nodes are possible emergency points aswell as potential base locations and the number of requiredambulances is small which is not always consistent withthe situation in Shanghai Therefore we further test the twoapproximated MIP formulations and the sampling approachon randomly generated instances by analysing the informa-tion on emergencies in Shanghai [26]
53 Computational Tests on Randomly Generated InstancesIn the following subsection numerical experiments on 120randomly generated instances are described
531 Test Instances By analysing the information on emer-gencies in Shanghai [26] the mean value 120583119895 of demand 119889119895 ineach emergency point are randomly generated fromadiscreteUniform distribution on interval [5 25] The standard devia-tion 120590119895 is set to be 5 Since the sampling approach proposed inNickel et al [8] requires a sample of scenarios in numericalexperiments scenarios are produced with demands randomlygenerated from Uniform distribution Poisson distributionand Normal distribution consistent with given mean andstandard deviation
According to the instance settings in Erkut and Ingolfsson(2008) response time from a base to an emergency pointis randomly generated from a discrete Uniform distributionon interval [3 30] in units of minutes A base covers anemergency point if the driving time between them is lessthan 10 minutes to ensure the survival probability (Erkut andIngolfsson 2006) For an emergency point in each instanceif there is no base with 10 minutes driving time to it then thebase covering this point is set to be the nearest one The costof constructing a base and that of locating one ambulance areboth set to be 1 as in Nickel et al [8] Besides the number ofemergency points is 3 times that of potential base locations
532 Results and Discussion Different values of requiredservice level 120572 are tested ie 085 09 095 Computationalresults on 120 randomly generated instances are reported inTables 4-6
Table 4 reports the numerical results of instances withdemand generated from Uniform distribution We observethat when 120572 = 085 09 095 the average computationaltimes of the sampling approach are 4482 4379 and 4295seconds respectively However both approximated MIP for-mulations can produce solutions in much faster speeds iewithin 13 seconds on average With the increase of 120572 (i)the out-of-sample average service level is getting higher and(ii) the system costs of solutions obtained by all method aregetting larger and (iii) the system costs of solutions yieldedbyMIP-DF2 is getting closer to those of solutions obtained bythe sampling approach In terms of the average system costthe sampling approach performs the best while MIP-DF1 isthe poorest When 120572 = 085 for example the average systemcost ofMIP-DF2 and the sampling approach are about 3032and 442 less than that of MIP-DF1 respectively When120572 = 095 the above gaps are even larger Concluding fortheUniformdistribution (1) the proposed two approximatedMIP formulations are far more time saving than the sampling
approach (2)MIP-DF1 andMIP-DF2 can improve the servicelevel by about 1907 compared with the sampling approachhowever (3) despite of the highest service levels obtainedby MIP-DF1 the system cost obtained by MIP-DF1 is alsohigh (4) MIP-DF2 can ensure an appropriate service levelwith much less cost than MIP-DF1 and (5) for the situationswith small-scale instances and high service level requirementMIP-DF1 is recommended
Computational results of instances with demand at eachemergency point generated from Poisson distribution arepresented in Table 5 For MIP-DF1 and MIP-DF2 onlymean and variance are involved Therefore with givenmean and standard deviation of the demand at each pointcomputational results are the same for different probabilitydistributions We can obtain from Table 5 that the objectivevalue obtained by the sampling approach increases with theincrease of 120572 Besides when the value of 120572 equals 085the average service level obtained by the sampling approachis about 8143 which is smaller than that of the twoapproximated MIP formulations The highest service level9934 is obtained by MIP-DF1 and that obtained by MIP-DF2 is 9708 When 120572 = 09 the average service levelof the sampling approach is 8437 and those obtained byMIP-DF1 and MIP-DF2 are 9985 and 9797 respectivelyWhen 120572 = 095 the average service level obtained by thesampling approach is 9356 and those yielded by MIP-DF1 and MIP-DF2 are 9997 and 9781 respectively ForPoisson distribution concluding MIP-DF1 and MIP-DF2can improve the service level compared with the samplingapproach especially MIP-DF1 can obtain the highest servicelevel
Similarly Table 6 reports the computational results forinstances in which the demand at each emergency point isgenerated following Normal distribution When 120572 = 085the average service levels obtained by the sampling approachMIP-DF1 and MIP-DF2 are 8174 9925 and 9742respectively When 120572 = 09 the three values are 84579981 and 9743 respectively Moreover for 120572 = 095 thevalues are equal to 8732 9999 and 9794 respectivelyIt can be observed that the performance of the samplingapproach is the poorest in terms of the service level
By observing the above computational results on bench-marks and randomly generated instances we conclude that(1)The proposed approximated MIP formulations ie MIP-DF1 and MIP-DF2 are very time saving compared with thesampling approach (2) MIP-DF1 an MIP-DF2 can achievethe requirement of service level and improve the service levelby about 1547 and 1229 on average compared with thatof the sampling approach (3) MIP-DF1 is more conservativein terms of overhigh emergency service level close to 100and the system cost is also high which is about 568 largerthan that of MIP-DF2 (4) Compared with MIP-DF1 MIP-DF2 is less conservative and much more cost saving whileensuring an appropriately service level on average which isabout 97
In sum for medium and large problem instances (egShanghai) which are common in practice to guarantee therequired service level with less cost we recommendMIP-DF2as solution method However for very small instances with
8 Journal of Advanced Transportation
Table4Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mun
iform
distrib
ution
Set (|119868||119869|)
120572=085
120572=09
120572=095
Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
1(26)
27
897317
01
149
9906
01
107
9222
28
967373
01
155
9933
01
107
9167
27
106
7585
01
185
9987
01
107
9128
2(412
)47
169
7423
00
291
9919
00
207
9122
58
181
7438
01
303
9997
01
207
9363
50
199
7702
00
363
9982
00
207
9338
3(618
)72
258
7506
01
438
9976
01
312
9106
75279
8082
01
456
9996
01
312
9369
78304
8337
01
546
9997
01
312
9357
4(824)
99314
7956
01
553
9993
01
385
9206
111
338
8202
01
577
9999
01
385
9602
112
366
8492
01
697
9994
01
385
9620
5(1030)
132
415
8406
01
722
9896
01
512
9406
143
441
8322
02
752
9983
01
512
9692
142
476
8609
01
902
9993
01
512
9692
6(1236)
170
475
8249
02
843
9870
01
591
9367
181
504
8251
01
879
9982
01
591
9608
189
535
8726
01
1059
9997
01
591
9609
7(14
42)
215
573
8111
02
1006
9849
02
712
9606
243
609
8217
02
1048
9953
02
712
9589
233
644
8608
02
1258
9992
02
712
9582
8(1648)
248
686
7532
02
1185
9896
02
849
9725
284
727
8307
02
1233
9949
02
849
9613
281
768
9063
02
1473
9999
03
849
9622
9(1854)
313
691
8353
03
1236
9974
04
858
9709
351
731
8397
03
1290
9988
03
858
9765
352
760
8906
03
1560
9994
03
858
9758
10(2060)
372
794
8174
03
1409
9887
03
989
9657
428
843
8489
03
1469
9947
03
989
9773
432
892
9004
03
1769
9997
03
989
9738
11(2266
)451
830
8072
04
1493
9949
03
1031
9758
518
880
8328
04
1559
9982
03
1031
9723
516
930
8967
05
1889
9995
03
1031
9727
12(2472)
543
1025
7970
04
1771
9927
04
1267
9835
601
1083
8482
05
1843
9949
04
1267
9776
637
1146
9014
04
2203
9993
04
1267
9777
13(2678)
631
1026
8142
05
1819
9898
04
1274
9782
743
1087
8678
05
1897
9936
04
1274
9778
765
1146
9023
04
2287
9997
05
1273
9837
14(2884)
782
1101
7903
05
1949
9869
08
1361
9804
878
1165
8790
05
2033
9935
07
1361
9889
870
1227
9056
05
2453
9998
06
1361
9808
15(3090)
907
1091
8098
06
1997
9915
06
1367
9629
1018
1155
8329
06
2087
9981
06
1367
9807
1020
1225
8986
06
2537
9999
08
1367
9817
16(3296)
1062
1160
8798
07
2123
9944
07
1449
9787
1183
1229
8383
07
2219
9988
06
1449
9849
1193
1300
9105
06
2698
9998
06
1449
9815
17(34102)
1202
1298
7572
07
2328
9969
07
1614
9821
1377
1374
8860
08
2430
9997
07
1614
9824
1260
1452
9085
07
2940
9999
08
1614
9818
18(36108)
1394
1408
8121
08
2509
9947
08
1752
9809
1544
1492
8928
08
2617
9983
08
1752
9799
1548
1580
9173
08
3156
9999
08
1752
9809
19(38114
)1618
1508
8170
09
2676
9899
111877
9742
1760
1598
9130
09
2790
9934
08
1877
9908
1755
1688
9186
09
3360
9999
09
1877
9872
20(40120)
2206
1628
7719
102869
9989
102029
9873
2007
1732
8753
112990
9997
09
2029
9879
1904
1827
9138
103590
9998
09
2029
9875
21(42126)
2780
1622
8354
102900
9978
102017
9732
2609
1718
8854
103026
9876
102017
9818
2670
1812
9089
123655
9999
102017
9848
22(44132)
3189
1735
8494
113088
9985
152165
9857
2970
1840
8922
123220
9925
122165
9886
2873
1952
9206
113880
9999
112165
9850
23(46138)
3630
1937
8141
123372
9893
132406
9796
3222
2048
8937
123510
9934
142406
9906
3461
2170
8989
124199
10000
122406
9809
24(48144)
4226
1968
8226
133461
9932
132453
9814
3674
2085
8846
133605
9956
152453
9912
3556
2211
9300
134326
9999
132453
9819
25(50150)
4727
1919
8111
143456
9999
152407
9629
4019
2037
8699
143606
9999
142407
9857
4257
2161
9259
154356
9995
142407
9819
26(52156)
5300
1906
8096
153484
9878
172392
9658
4495
2021
8952
143640
9999
152392
9568
4486
2142
9182
1544
209999
152392
9854
27(54162)
5913
2096
8387
163751
9846
172617
9761
5117
2221
9045
163913
9998
172617
9544
4830
2355
9346
184723
9999
162617
9893
28(56168)
6310
2110
7758
173819
9933
192643
9829
5777
2234
9157
183987
9998
182643
9526
5550
2365
9045
184827
9999
172643
9831
29(58174)
6780
2175
7998
183937
9898
182720
9714
6303
2307
9022
20
4111
9969
192720
9623
6189
2444
9128
194981
10000
192720
9914
30(60180)
7593
2384
8321
194240
9967
20
2979
9841
6791
2527
8978
1944
199979
23
2979
9721
6878
2679
9143
21
5316
9999
20
2979
9869
31(62186)
8220
2281
7604
20
4157
9984
21
2856
9796
7734
2416
8934
21
4343
9997
21
2856
9679
7466
2557
9158
22
5273
9999
23
2856
9902
32(64192)
8766
2517
8378
21
4489
9932
25
3145
9729
8276
2671
8296
23
4680
9977
24
3145
9825
7897
2822
9089
24
5640
9999
22
3145
9891
33(66198)
10123
2532
7806
23
4555
9967
23
3169
9615
9010
2682
8947
24
4752
9989
24
3169
9703
9051
2835
9157
24
5742
9998
24
3169
9867
34(68204)
11010
2535
7934
24
4602
9999
29
3174
9631
9898
2685
8858
30
4806
9998
28
3174
9762
9863
2844
9236
26
5824
9999
25
3174
9889
35(70210)
11666
2714
8062
25
4862
9986
29
3391
9787
1066
62875
9137
26
5072
9999
29
3391
9716
10822
3043
9315
26
6122
9999
29
3391
9923
36(72216)
12011
2827
8104
26
5047
9991
28
3535
9798
1246
52984
9016
28
5263
9999
28
3535
9878
11667
3169
9253
30
6343
9999
29
3535
9896
37(74222)
11437
2826
8228
28
5096
9933
29
3543
9884
13582
3098
8895
29
5318
9999
32
3543
9888
12654
3178
9139
29
6427
10000
31
3543
9893
38(76228)
12435
2839
8152
29
5161
9897
29
3565
9788
13976
3103
8926
29
5388
9999
31
3565
9724
13837
3185
9255
33
6528
10000
31
3565
9895
39(78234)
1464
32945
8412
31
5331
9998
32
3695
9876
14794
3210
9005
31
5565
9999
34
3695
9713
14561
3313
9081
34
6735
10000
33
3695
9878
40(80240)
1604
22975
8510
32
5414
9957
36
3734
9796
16263
3346
8999
35
5654
9999
35
3734
9806
15869
3343
9248
35
6854
10000
35
3734
9873
Average
4482
1584
680
67
1228
397
9936
131978
796
82
4379
16913
8654
1329
626
9975
131978
79721
4295
1778
889
85
1335774
9997
131978
79775
Journal of Advanced Transportation 9
Table5Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mPo
isson
distr
ibution
set ( |119868 |
|119869 |)120572=
085120572=
09120572=
095Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
1(26)
27
907294
9967
9001
27
987770
9979
9013
24
103
9273
9999
9468
2(412
)46
170
7961
9921
9117
51
183
8305
9946
9128
49
201
9392
9998
9546
3(618
)68
258
7578
9938
8911
74280
8028
9968
9004
69
307
9182
9999
9658
4(824)
93315
8001
9928
9196
102
338
8438
9967
9172
92368
9468
9999
9613
5(1030)
120
413
8041
9939
9449
133
443
8447
9976
9272
121
479
9544
9999
9752
6(1236)
155
476
8084
9877
9554
167
508
8863
9991
9569
159
536
9456
9999
9734
7(14
42)
195
573
7876
9892
9663
213
609
9124
9998
9729
205
643
9314
10000
9768
8(1648)
232
687
7964
9997
9788
267
724
8523
9988
9687
238
772
9422
9998
9689
9(1854)
293
692
8267
9893
9809
327
730
8386
9999
9750
298
757
9228
9999
9755
10(2060)
356
796
7946
9899
9801
399
840
8299
9976
9686
360
885
9668
9999
9705
11(2266)
423
829
8201
9895
9756
489
877
8378
9987
9788
448
927
9476
9999
9812
12(2472)
512
1023
8338
9913
9853
554
1085
8501
9999
9881
514
1148
9398
9999
9888
13(2678)
592
1045
8416
9907
9873
687
1088
7998
9996
9802
607
1152
9504
9999
9821
14(2884)
723
1099
7946
9915
9788
838
1166
8056
9992
9829
776
1230
9546
9999
9839
15(3090)
855
1089
7915
9964
9703
942
1156
8432
9984
9778
903
1225
9223
9998
9785
16(3296)
1009
1158
8024
9897
9769
1163
1231
8500
9996
9824
1045
1222
9199
9999
9835
17(34102)
1165
1299
8034
9919
9779
1285
1375
8495
9937
9808
1190
1457
9258
10000
9846
18(36108)
1254
1411
8176
9943
9885
1479
1492
8561
9968
9868
1345
1579
9286
9999
9875
19(38114
)1504
1506
7970
9923
9842
1651
1594
8167
9972
9846
1564
1694
9376
9999
9786
20(40120)
1706
1635
7819
9916
9783
1854
1734
8607
9968
9776
1778
1825
9285
9904
9718
21(42126)
2054
1622
8354
9918
9834
2357
1720
8297
9978
9833
2226
1856
9501
9999
9834
22(44132)
2400
1736
8494
9922
9846
2852
1844
8269
9999
9816
2680
1946
9486
9999
9820
23(46138)
2651
1934
8141
9953
9793
3159
2049
8492
9969
9799
3072
2172
9358
10000
9813
24(48144)
3004
1961
8226
9947
9681
3522
2086
8513
9998
9706
3102
2209
9488
9999
9724
25(50150)
3308
1924
8111
9965
9746
3898
2038
8264
9999
9788
3639
2161
9546
9999
9813
26(52156)
3669
1911
8096
9939
9855
4305
2027
8722
9967
9874
4453
2142
9185
9999
9876
27(54162)
4418
2095
8387
9927
9754
4771
2219
8341
9989
9749
4806
2353
9289
9999
9755
28(56168)
5367
2113
8458
9968
9848
5356
2237
8452
9978
9855
5221
2369
9228
10000
9868
29(58174)
6217
2186
7998
9989
9694
5942
2315
8316
9996
9699
5853
2371
9167
9999
9705
30(60180)
7263
2382
8321
9991
9837
6527
2543
8359
9998
9886
6214
2678
9369
9997
9886
31(62186)
8214
2346
8404
9936
9888
7813
2430
8321
9999
9846
6876
2560
9358
9999
9847
32(64192)
8863
2518
8378
9938
9807
8106
2677
8356
9999
9833
7877
2822
9408
10000
9835
33(66198)
9985
2532
8406
9917
9709
9684
2689
8497
9996
9746
8088
2838
9507
9999
9749
34(68204)
11264
2534
8543
9908
9897
10986
2685
8809
9996
9782
8469
2843
8844
9998
9780
35(70210)
12035
2715
8258
9937
9799
12098
2886
8569
9991
9809
9946
3041
9459
9999
9813
36(72216)
12803
2825
8104
9929
9827
13076
3010
8651
9994
9834
11013
3170
9468
9999
9865
37(74222)
13021
2830
8228
9949
9826
13261
3009
8376
9999
9826
12014
3172
9346
9999
9888
38(76228)
13862
2843
8152
9933
9807
13995
3028
8760
9998
9841
12133
3187
9706
9999
9876
39(78234)
15387
2953
8312
9964
9799
14827
3133
9001
9995
9783
13381
3304
8913
9999
9799
40(80240)
16974
2975
8510
9985
9767
15256
3249
8217
9993
9781
15867
3347
9111
9999
9791
Average
4352
15875
8143
9934
9708
4362
16856
8437
9985
9707
3968
17763
9356
9997
9781
10 Journal of Advanced Transportation
Table6Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mno
rmaldistr
ibution
set ( |119868 |
|119869 |)120572=
085120572=
09120572=
095Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
1(26)
27
907294
9964
9117
27
947844
9970
9202
2558
106
8422
10000
9208
2(412
)46
170
7961
9921
9283
48
183
8328
9982
9283
4508
183
8823
10000
9370
3(618
)68
258
7578
9916
9228
68
280
7798
9971
9417
6803
307
7658
9998
9447
4(824)
93315
8000
9899
9367
96337
8379
9979
9467
9373
368
8438
9999
9579
5(1030)
120
413
7841
9927
9437
124
443
8763
9963
9476
1262
499
8850
9998
9708
6(1236)
155
476
7921
9919
9519
162
504
8124
9982
9582
16021
536
8954
10000
9756
7(14
42)
195
573
7976
9933
9598
196
609
8287
9975
9619
2011
646
8467
9999
9767
8(1648)
232
687
8031
9918
9705
235
727
8115
9965
9650
2404
772
8278
9999
9765
9(1854)
293
692
8086
9898
9759
291
734
8464
9974
9755
29815
796
8617
9998
9835
10(2060)
356
796
7984
9937
9770
358
842
8223
9988
9765
37973
921
8846
10000
9838
11(2266
)421
829
8101
9940
9755
436
877
8211
9986
9786
4514
8959
8497
9999
9789
12(2472)
514
1023
8033
9951
9785
501
1085
8162
9983
9788
52073
1147
8321
9999
9824
13(2678)
592
1023
8138
9916
9885
599
1087
8411
9970
9845
62937
1194
8609
9998
9795
14(2884)
728
1099
8235
9913
9915
722
1168
8435
9987
9806
75053
1270
8567
9997
9834
15(3090)
855
1089
8187
9922
9861
880
1157
8376
9976
9796
89222
1300
8611
9996
9798
16(3296)
1009
1158
8770
9897
9788
966
1231
8845
9984
9823
106288
1359
8997
9999
9848
17(34102)
1154
1299
8434
9925
9849
1137
1377
8601
9980
9798
123869
1493
8798
9996
9877
18(36108)
1407
1411
8321
9927
9866
1339
1495
8301
9965
9822
133807
1601
8567
9999
9834
19(38114
)1702
1506
8206
9915
9805
1672
1598
8977
9978
9845
152999
1694
8998
9999
9847
20(40120)
2091
1635
8528
9918
9889
1916
1737
8805
9976
9837
1890
021833
9012
9999
9791
21(42126)
2843
1622
8550
9934
9913
2432
1718
8886
9974
9813
2399
131815
9008
10000
9815
22(44132)
3300
1736
8449
9940
9859
2795
1843
8786
9977
9756
253106
1948
8997
9999
9826
23(46138)
3741
1934
8133
9911
9876
3143
2050
8876
9981
9785
274583
2173
9014
9999
9858
24(48144)
4531
1961
8245
9909
9846
3515
2090
8344
9993
9819
308089
2211
8661
9999
9883
25(50150)
4715
1924
8111
9929
9888
3959
2041
8279
9996
9837
3510
932158
8527
9999
9827
26(52156)
5223
1911
8096
9898
9857
4301
2025
8251
9997
9757
421747
2146
8497
9999
9914
27(54162)
5994
2095
8387
9908
9775
4835
2226
8497
9974
9780
4614
432359
8695
9999
9837
28(56168)
6406
2111
8411
9906
9798
5336
2240
8558
9991
9820
494113
2371
9249
9999
9816
29(58174)
6897
2176
8012
9916
9791
5948
2311
8211
9984
9765
624835
2445
8531
10000
9851
30(60180)
7436
2391
8111
9913
9766
6558
2533
8305
9968
9795
696116
2682
8575
9999
9799
31(62186)
8118
2282
8119
9928
9806
7069
2422
8417
9973
9873
797768
2564
8667
9999
9846
32(64192)
8895
2519
8223
9919
9757
7857
2675
8631
9982
9769
864813
2827
8976
10000
9905
33(66198)
11235
2533
8563
9928
9811
9162
2688
8547
9979
9828
973089
2843
8657
9999
9846
34(68204)
11587
2534
8249
9926
9847
9549
2686
8798
9938
9856
10523
3041
8999
10000
9877
35(70210)
1164
92715
8098
9895
9872
10345
2880
8602
9998
9869
11272
3052
9021
9999
9897
36(72216)
12224
2825
8403
9958
9770
11165
3002
8657
9999
9839
12456
3175
9113
9999
9864
37(74222)
12978
2830
8357
9937
9795
11850
3005
8496
9998
9894
13458
3176
8827
9999
9905
38(76228)
13426
2844
8267
9966
9759
12795
3019
8604
9999
9797
15155
3121
8978
9999
9853
39(78234)
15003
2955
8303
9956
9867
13816
3132
8588
9996
9885
15041
3310
9016
9999
9912
40(80240)
16495
2978
8257
9954
9826
14896
3160
8495
9997
9837
16563
3349
8957
9999
9933
Average
4619
15855
8174
9925
9742
4077
16828
8457
9981
9743
4350
17938
8732
9999
9794
Journal of Advanced Transportation 11
high service level requirement we recommend MIP-DF1 forsolutions
6 Conclusion
This paper investigates an ambulance location problem withambiguity set of demand in which the probability distribu-tion of the uncertain demand is unknown In such a data-driven environment only the mean and covariance demandsat emergency points are knownThe problem is to determinethe base locations and the number of ambulances at eachbase aiming at minimizing the total cost associated withlocating bases and assigning ambulances We propose adistribution-free model with chance constraints Then twoapproximated MIP formulations with different approxima-tion methods of chance constraints are proposed which arebased on different ambiguity sets of the unknown probabilitydistribution Numerical experiments on benchmarks and120 randomly generated instances are conducted and thecomputational results show that the first approximated MIPformulation is much conservative with overhigh system costwhile the second approximatedMIP formulation ismore costsaving with service level appropriately ensured
In future research wemay consider the following relevantissues First the uncertainty of driving time from a base toan emergency point shall be analysed Besides unexpected orsudden disasters may be taken into consideration Moreoverother uncertainties should be considered such as travel timethe number of available vehicles and so on The stochasticdependence should also further considered
Data Availability
The data of the benchmark instances in our manuscript canbe obtained in Nickel et al [8] For the test instances thedata is randomly generated and the way for generating data isstated in our manuscript The datasets generated during thecurrent study are available from the corresponding author onreasonable request
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This work was supported by the National Natural Sci-ence Foundation of China (NSFC) under Grants 7142800271531011 71571134 and 71771048 This work was also sup-ported by the Fundamental Research Funds for the CentralUniversities
References
[1] L Brotcorne G Laporte and F Semet ldquoAmbulance loca-tion and relocation modelsrdquo European Journal of OperationalResearch vol 147 no 3 pp 451ndash463 2003
[2] X Li Z Zhao X Zhu and T Wyatt ldquoCovering modelsand optimization techniques for emergency response facilitylocation and planning a reviewrdquo Mathematical Methods ofOperations Research vol 74 no 3 pp 281ndash310 2011
[3] C Toregas R Swain C ReVelle and L Bergman ldquoThe locationof emergency service facilitiesrdquoOperations Research vol 19 no6 pp 1363ndash1373 1971
[4] R Church and C ReVelle ldquoThe maximal covering locationproblemrdquo Papers in Regional Science vol 32 no 1 pp 101ndash1181974
[5] M Gendreau G Laporte and F Semet ldquoSolving an ambulancelocation model by tabu searchrdquo Location Science vol 5 no 2pp 75ndash88 1997
[6] A Ingolfsson S Budge and E Erkut ldquoOptimal ambulancelocation with random delays and travel timesrdquo Health CareManagement Science vol 11 no 3 pp 262ndash274 2008
[7] M Gendreau G Laporte and F Semet ldquoThemaximal expectedcoverage relocation problem for emergency vehiclesrdquo Journal ofthe Operational Research Society vol 57 no 1 pp 22ndash28 2006
[8] S Nickel M Reuter-Oppermann and F Saldanha-da-GamaldquoAmbulance location under stochastic demand A samplingapproachrdquo Operations Research for Health Care vol 8 pp 24ndash32 2016
[9] P Beraldi M E Bruni and D Conforti ldquoDesigning robustemergency medical service via stochastic programmingrdquo Euro-pean Journal of Operational Research vol 158 no 1 pp 183ndash1932004
[10] N Noyan ldquoAlternate risk measures for emergency medicalservice system designrdquo Annals of Operations Research vol 181pp 559ndash589 2010
[11] M R Wagner ldquoStochastic 0-1 linear programming underlimited distributional informationrdquoOperations Research Lettersvol 36 no 2 pp 150ndash156 2008
[12] EDelage andYYe ldquoDistributionally robust optimization undermoment uncertainty with application to data-driven problemsrdquoOperations Research vol 58 no 3 pp 595ndash612 2010
[13] J T Essen J L Hurink S Nickel and M Reuter ldquoModels forambulance planning on the strategic and the tactical levelrdquo BetaResearch School for Operations Management and Logistics p 272013
[14] E Erkut A Ingolfsson and G Erdogan ldquoAmbulance locationfor maximum survivalrdquoNaval Research Logistics (NRL) vol 55no 1 pp 42ndash58 2008
[15] S C Chapman and J A White ldquoProbabilistic formulations ofemergency service facilities location problemsrdquo in Proceedingsof ORSATIMS Conference San Juan Puerto Rico 1974
[16] A Ben-Tal D Den Hertog A De Waegenaere B Melenbergand G Rennen ldquoRobust solutions of optimization problemsaffected by uncertain probabilitiesrdquo Management Science vol59 no 2 pp 341ndash357 2013
[17] MNg ldquoDistribution-free vessel deployment for liner shippingrdquoEuropean Journal of Operational Research vol 238 no 3 pp858ndash862 2014
[18] M Ng ldquoContainer vessel fleet deployment for liner shippingwith stochastic dependencies in shipping demandrdquo Transporta-tion Research Part B Methodological vol 74 pp 79ndash87 2015
[19] R Jiang and Y Guan ldquoData-driven chance constrained stochas-tic programrdquo Mathematical Programming vol 158 no 1-2 SerA pp 291ndash327 2016
[20] C-M Lee and S-L Hsu ldquoThe effect of advertising on thedistribution-free newsboy problemrdquo International Journal ofProduction Economics vol 129 no 1 pp 217ndash224 2011
[21] K Kwon and T Cheong ldquoA minimax distribution-free proce-dure for a newsvendor problem with free shippingrdquo EuropeanJournal of Operational Research vol 232 no 1 pp 234ndash2402014
12 Journal of Advanced Transportation
[22] Y Zhang S Shen and S A Erdogan ldquoDistributionally robustappointment scheduling with moment-based ambiguity setrdquoOperations Research Letters vol 45 no 2 pp 139ndash144 2017
[23] Y Zhang S Shen and S A Erdogan ldquoSolving 0-1semidefinite programs for distributionally robust allocationof surgery blocksrdquo Social Science Electronic Publishing 2017httpsssrncomabstract=2997524
[24] F Zheng J He F Chu and M Liu ldquoA new distribution-freemodel for disassembly line balancing problem with stochastictask processing timesrdquo International Journal of ProductionResearch pp 1ndash13 2018
[25] J Cheng E Delage and A Lisser ldquoDistributionally robuststochastic knapsack problemrdquo SIAM Journal on Optimizationvol 24 no 3 pp 1485ndash1506 2014
[26] M Liu D Yang and F Hao ldquoOptimization for the locationsof ambulances under two-stage life rescue in the emergencymedical service a case study in Shanghai ChinardquoMathematicalProblems in Engineering vol 2017 Article ID 1830480 14 pages2017
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
Journal of Advanced Transportation 3
of confidence sets for the possible probability distributionsare proposed For more distribution-free formulation appli-cations please see Lee and Hsu [20] Kwon and Cheong [21]etc for the stochastic inventory problem and Zhang et al[23] for stochastic allocating surgeries problem in operatingrooms and Zheng et al [24] for stochastic disassembly linebalancing problem To the best of our knowledge there is noresearch for the stochastic ambulance location problem withonly partial information on the uncertain demand
3 Problem Description and Formulation
In this section we first describe the considered problem andthen propose a new distribution-free model
31 Problem Description There is a given set of candidatebase locations 119868 = 1 2 |119868| and a set of potentialemergency demand points 119869 = 1 2 |119869| The emergencypoints may refer to a road a part of the urban area and avillage Base 119894 isin 119868 is said to be covering an emergency point119895 if the driving time 119905119894119895 between 119894 and 119895 is no more than apredetermined value119879The set 119868119895 of candidate bases coveringemergency point 119895 is denoted as 119868119895 = 119894 isin 119868 | 119905119894119895 le 119879
The number of ambulances required by an emergencypoint 119895 isin 119869 is denoted as 119889119895 which depends on the severityof the practical situationWe consider uncertain emergencieswhich are estimated or predicted by partial distributionalinformation thus the demand is regarded to be ambiguousBesides we focus on the case where the historical data cannotbe represented by a precise probability distribution It isassumed that the customer demands are independent of eachother
The objective of the problem is to select a subset ofbase locations and determine the number of ambulances ateach constructed base in order to minimize the total baseconstruction cost and ambulance costThroughout this paperwe assume that only the mean and covariance matrix ofdemands are known Moreover a predefined safety level 120572119895is given for each potential emergency point 119895 isin 119869 That is thenumber of ambulances serving each emergency point 119895 isin 119869 islarger than or equal to its demand with a least probability of12057211989532 Chance Constraint Construction In this section weintroduce a chance constraint to guarantee the safety levelof each emergency demand point In the following 119910119894119895 is adecision variable denoting the number of ambulances servingpoint 119895 isin 119869 from base 119894 isin 119868119895 and 119889119895 is the uncertain demandat point 119895 isin 119869 The chance constrained inequality is presentedas follows
ProbP (sum119894isin119868119895
119910119894119895 ge 119889119895) ge 120572119895 forall119895 isin 119869 (1)
where ProbP(sdot) denotes the probability of the event inparentheses under any potential probability distribution PConstraint (1) ensures that the number of ambulances servingemergency point 119895 isin 119869 is no less than its demand with a leastprobability of 120572119895 (ie safety level)
33 Distribution-Free Formulation In the following we givebasic notations define decision variables and propose thedistribution-free formulationDF for the ambulance locationproblem with uncertain demand
Parameters
119894 index of base locations119895 index of emergency points119868 set of candidate base locations119869 set of possible emergency demand points119905119894119895 driving time between base 119894 isin 119868 and demand point119895 isin 119869119879 maximum driving time for serving any emergencycall119868119895 set of candidate bases that can cover emergencypoint 119895 isin 119869 ie 119868119895 = 119894 isin 119868 | 119905119894119895 le 119879119869119894 set of potential emergency points covered by base119894 isin 119868 ie 119869119894 = 119895 isin 119869 | 119905119894119895 le 119879119891119894 fixed construction cost for installing a base atlocation 119894 isin 119868119892119894 fixed cost associated with an ambulance to belocated at 119894 isin 119868119889119895 number of (stochastic) ambulances requested byemergency point 119895 isin 119869M a sufficiently large positive number
Decision Variables
119909119894 a binary variable equal to 1 if a base is constructedat 119894 isin 119868 0 otherwise119911119894 number of ambulances assigned to possible baselocation 119894 isin 119868119910119894119895 number of ambulances serving emergency point119895 isin 119869 from possible base location 119894 isin 119868119895
Distribution-Free Model [DF]
[DF]
min sum119894isin119868
(119891119894 sdot 119909119894 + 119892119894 sdot 119911119894)st Constraint (1)
(2)
119911119894 le M sdot 119909119894 forall119894 isin 119868 (3)
sum119895isin119869119894
119910119894119895 le 119911119894 forall119894 isin 119868 (4)
119909119894 isin 0 1 forall119894 isin 119868 (5)
119911119894 isin Z+ forall119894 isin 119868 (6)
119910119894119895 isin Z+ forall119894 isin 119868 119895 isin 119869119894 (7)
4 Journal of Advanced Transportation
The objective function denotes the goal to minimize thetotal cost consisting of two parts (i) the cost for constructingbases ie sum119894isin119868 119891119894 sdot 119909119894 and (ii) the cost for deploying ambu-lances ie sum119894isin119868 119892119894 sdot 119911119894
Constraint (3) ensures that ambulances can only belocated at the opened bases Constraint (4) ensures that thenumber of ambulances sent to serve emergency points frombase 119894 isin 119868 does not exceed the total number of ambulanceslocated at 119894 isin 119868 Constraints (5)-(7) are the restrictions ondecision variables
4 Solution Approaches
The proposed distribution-free model is difficult to solvewith the commercial software due to chance constraints Inthis section we propose two approximatedMIP formulationsbased on those in Wagner [11] and in Delage and Ye [12]respectively In the following subsections we present the twoapproximated MIP formulations
41 Approximated MIP Formulation MIP-DF1 In this partwe first describe a widely used ambiguity set to describe theuncertainty Then an approximated MIP formulation MIP-DF1 is proposed
Given a set of independent historical data samples 119889119904|119878|119904=1of random vectors of demands where 119878 is the set of sampleindexes the mean vector 120583 and the covariance matrix Σ ofdemands can be estimated as follows
120583 = 1|119878|sum119904isin119878119889119904Σ = 1|119878|sum119904isin119878 (119889119904 minus 120583) (119889119904 minus 120583)⊤
(8)
where (sdot)⊤ implies the transposition of the vector in paren-theses Then an ambiguity set of all probability distributionsof demandsP1(120583 Σ) can be described as the following
P1 (120583 Σ) = P EP [119889] = 120583EP [(119889 minus 120583) (119889 minus 120583)⊤] = Σ (9)
whereP denotes a possible probability distribution satisfyingthe given conditions and EP[sdot] denotes the expected valueof the number in parentheses Then the chance constraint (1)with the ambiguity setP1 can be presented as follows
ProbP (sum119894isin119868119895
119910119894119895 ge 119889119895) ge 120572119895 forall119895 isin 119869 P isin P1 (10)
According to Wagner (2010) and Ng [18] constraint (1) canbe conservatively approximated by the following
sum119894isin119868119895
119910119894119895 ge 120583119895 + 120590119895 sdot radic 120572119895(1 minus 120572119895) forall119895 isin 119869 (11)
where 120590119895 = radicΣ119895119895 and Σ119895119895 denotes the 119895-th element in the 119895-thcolumn of matrix Σ In terms of conservative approximation
all solutions satisfying constraint (11) must satisfy constraint(1) Then we describe the approximated MIP formulationMIP-DF1 toDF combined with constraint (11)
[MIP-DF1]
min sum119894isin119868
(119891119894 sdot 119909119894 + 119892119894 sdot 119911119894)st Constraints (4) - (8) (12)
(12)
MIP-DF1 can be exactly solved by the commercialsoftware such as CPLEX As we can observe from thecomputational results in Section 5 it can obtain a relativelyhigh service level However the system cost is also high Inorder to save the system cost another approximated MIPformulation of DF is then proposed in the next subsection
42 Approximated MIP Formulation MIP-DF2 AmbiguitysetP1 focuses on exactly matching the mean and covariancematrix of uncertain parameters However in practice theremay be considerable estimation errors in the mean andcovariance matrix To take the inevitable estimation errorsinto consideration Delage and Ye [12] introduce a newmoment-based ambiguity set Therefore we in the followingemploy the moment-based ambiguity set
P2 (120583 Σ 1205741 1205742)= P (EP [119889] minus 120583)⊤ Σminus1 (EP [119889] minus 120583) le 1205741
EP [(119889 minus 120583) (119889 minus 120583)⊤] ⪯ 1205742Σ (13)
where 1205741 ge 0 and 1205742 ge 0 are two parameters of ambiguityset P2(120583 Σ 1205741 1205742) with the following assumptions (i) thetrue mean vector of demands is within an ellipsoid of sizeproportion to 1205741 centered at 120583 and (ii) the true covariancematrix of demands is in a positive semidefinite cone boundedby amatrix inequality of 1205742ΣThe ambiguity set describes howlikely the uncertain parameters are to be close to the meanin terms of the correlation [12] Besides under the moment-based ambiguity set various stochastic programs with partialdistributional information have been successfully modeledand solved [22 23 25]
According to the approximation method in Zhang et al[23] chance constraint (1) can be approximated by
radic 11 minus 119886 minus 119887 sdot (1 + radic1 minus 120572119895120572119895 sdot 119887) sdot 120590119895le radic1 minus 120572119895120572119895 sdot (sum
119894isin119868119895
119910119894119895 minus 120583119895) 119895 isin 119869(14)
where
119886 = 1 minus 1205741 + 11205742 minus 1205741 119887 = 12057411205742 minus 1205741
(15)
Journal of Advanced Transportation 5
Table 1 120572 = 095|119868| = |119869| Sampling approach MIP-DF1 MIP-DF2
CT Obj Sel () CT Obj Sel () CT Obj Sel ()1 7 49 23 9626 01 36 10000 01 23 100002 7 51 22 9863 01 35 10000 01 24 99823 6 41 14 9692 01 30 10000 01 14 98504 6 43 13 9713 02 29 10000 01 12 97425 6 57 12 7996 01 30 10000 02 12 100006 8 58 21 9850 01 24 10000 02 27 10000
Average 50 175 9457 01 307 10000 01 187 9929
The second approximated MIP formulationMIP-DF2 toDF is presented as follows
[MIP-DF2]
min sum119894isin119868
(119891119894 sdot 119909119894 + 119892119894 sdot 119911119894)st Constraints (4) - (8) (15)
(16)
Notice that when 1205741 = 0 and 1205742 = 1 inequality (14)reduces to (11) in MIP-DF1 which implies that MIP-DF1 isa special case of MIP-DF2
5 Computational Experiments
In this section the performance of our proposed formula-tions first evaluated and compared to the sampling approachproposed in Nickel et al [8] Specifically given a requiredservice level 120572 we apply our MIP formulations by restrictingthe chance constraint for each emergency point to satisfya safety level 120572119895 = 120572 while we adopt the samplingapproach by restricting the coverage constraint with 120572 Thencomputational results on 120 randomly generated instancesbased on the information on emergencies in Shanghai arereported Our approximated MIP formulations and thesampling approach are solved by coding in MATLAB 2014bcalling CPLEX 126 solver All numerical experiments areconducted on a personal computer with Core I5 and 330GHzprocessor and 8GB RAMunder windows 7 operating systemThe computational time for sampling approach is limited to3600 seconds
51 Out-of-Sample Test FollowingZhang et alrsquos [22]method-ology we examine the out-of-sample performance of solu-tions obtained by our proposed approximated MIP formula-tions and the sampling approach [8] The out-of-sample testincludes three steps
(1) The base locations and the employment of ambu-lances at each base are determined with known meanvector and covariance matrix of demands which isnamed as in-sample test
(2) 1000 scenarios are sampled from the underlying truedistribution which represent realizing the uncertaindemand
(3) Based on the solution obtained in step (1) ambu-lances are assigned to serve emergency points inthe 1000 scenarios to maximize the (out-of-sample)service level
The (out-of-sample) service level is defined as the ratioof the number of emergency points with demand satisfiedto the total number of emergency points in 1000 scenariosIn the following subsections the benchmarks are tested andthen three common distributions are employed ie uniformdistribution Poisson distribution and normal distributionrespectively
52 Benchmark Instances We first test the benchmarks inNickel et al [8] to compare our proposed two approximatedMIP formulationswith the sampling approach InNickel et al[8] they assume that demand nodes are also potential baselocations and the number of ambulances required by eachemergency point is chosen from set 0 1 2with given proba-bility distribution Computational results on benchmarks arereported in Tables 1 and 2
Table 1 reports the computational results with servicelevel requirement120572 = 095 In Table 1 CTObj and Sel denotethe computational time objective value (ie the system cost)and the service level respectively The first column includesthe indexes of instances and the second indicates the numberof nodes It shows that the average service level obtainedby the sampling approach is 9457 which is smaller thanthe required service level ie 85 Specifically the servicelevel of instance 5 obtained by the sampling approach is only7996 However MIP-DF1 and MIP-DF2 can guarantee theservice levels for all instances higher than the requirementConcluding from Table 1 we can observe that (i) the averagecomputational time of sampling approach is about 50 timesas those of the two approximated MIP formulations and (ii)MIP-DF1 and MIP-DF2 improve the average service level byabout 574 and 499 compared with sampling approachhowever (iii) system costs obtained by MIP-DF1 and MIP-DF2 are respectively 7543 and 686 higher than that ofsampling approach
Table 2 reports the computational results with servicelevel requirement 120572 = 099 For instances 2-6 the ser-vice levels obtained by sampling approach cannot meet therequirement ie 99 Besides it shows that the averagecomputational time of sampling approach is 48 which isabout 48 times greater than those of MIP-DF1 andMIP-DF2
6 Journal of Advanced Transportation
Table 2 120572 = 099|119868| = |119869| Sampling approach MIP-DF1 MIP-DF2
CT Obj Sel () CT Obj Sel () CT Obj Sel ()1 7 49 24 9979 01 64 10000 01 25 100002 7 48 23 9856 01 63 10000 01 24 100003 6 51 15 9717 01 55 10000 01 16 100004 6 47 15 9275 01 52 10000 01 13 100005 6 40 14 9000 01 55 10000 01 14 100006 8 52 23 9785 01 45 10000 01 27 10000
Average 48 190 9343 01 557 10000 01 198 10000
Table 3 Solutions to instance 4 in Nickel et al [8] for the given scenario
Sampling approach MIP-DF1 MIP-DF2Emergency points 1 2 3 4 5 6 Emergency points 1 2 3 4 5 6 Emergency points 1 2 3 4 5 6
Demand 1 0 2 0 2 0 Demand 1 0 2 0 2 0 Demand 1 0 2 0 2 0Base locations Ambulance employment Ambulance employment Ambulance employment1 0 0 02 1 1 0 3 1 23 1 1 26 1 2 2 04 0 0 05 0 6 06 2 2 0 4 2
The average service level is 9343 Although the averageservice levels obtained by MIP-DF1 and MIP-DF2 are both100 the average system cost obtained by MIP-DF1 is about3 times larger than that of sampling approach Besides theaverage system cost obtained by MIP-DF2 is 421 higherthan that of sampling approach
Concluding for the benchmarks described in Nickel etal [8] we can observe that (1) MIP-DF1 and MIP-DF2 canbe solved in far less computational time compared with thesampling approach (2) the sampling approach cannot meetthe requirement of service level for all benchmarks whilethe two MIP formulations can (3) the average system cost ofMIP-DF2 is very close to that of sampling approach (4) withthe increase of 120572 the average system costs obtained by MIP-DF2 is getting closer to that obtained by sampling approach
From Tables 1 and 2 we can observe that in spite ofthe lower system cost obtained by the sampling approachthere may exist extreme situations that demands of someemergency points are not satisfied Take one benchmark inNickel et al [8] ie instance 4 where there are 6 nodesA scenario in which demand at each emergency point isgenerated from the given probability distribution is shown inFigure 1 We can observe from Figure 1 that under the givenscenario the demand of each possible emergency point is1 0 2 0 2 0The solutions obtained by sampling approachMIP-DF1 andMIP-DF2 are shown in Table 3 which includesbase locations and ambulance employment obtained by in-sample test and ambulance assignment under given scenario
In Table 3 the first column includes the indexes of poten-tial base locations The second column represents the baselocations and the ambulance employment at each location
>1 = 1
>2 = 0
>3 = 2
>4 = 0
>5 = 2
>6 = 0
1
2 4
3 5
6
Figure 1 Instance 4 in Nickel et al [8] under given scenario
in which a location with 0 ambulance implies that it is notselected for base construction From Table 3 we can observethat the in-sample solutions obtained by sampling approachis to construct bases at locations 2 3 and 6 where the numberof ambulances is 1 1 and 2 respectively Besides undergiven scenario there is one ambulance serving emergencypoint 1 from base location 3 and one ambulance servingemergency point from base location 2 and 2 ambulancesserving emergency point 5 from base location 6 It shows thatthe requirement of ambulances at emergency point 3 is notfully satisfied The insufficient service may cause serious lossof life However we can find that the solutions obtained byMIP-DF1 andMIP-DF2 can provide sufficient ambulances forall emergency points
In sum to guarantee the required service level withless cost MIP-DF2 is recommended for the solutions with
Journal of Advanced Transportation 7
Nickel et al [8]rsquos data set Moreover in the benchmarks itis assumed that all nodes are possible emergency points aswell as potential base locations and the number of requiredambulances is small which is not always consistent withthe situation in Shanghai Therefore we further test the twoapproximated MIP formulations and the sampling approachon randomly generated instances by analysing the informa-tion on emergencies in Shanghai [26]
53 Computational Tests on Randomly Generated InstancesIn the following subsection numerical experiments on 120randomly generated instances are described
531 Test Instances By analysing the information on emer-gencies in Shanghai [26] the mean value 120583119895 of demand 119889119895 ineach emergency point are randomly generated fromadiscreteUniform distribution on interval [5 25] The standard devia-tion 120590119895 is set to be 5 Since the sampling approach proposed inNickel et al [8] requires a sample of scenarios in numericalexperiments scenarios are produced with demands randomlygenerated from Uniform distribution Poisson distributionand Normal distribution consistent with given mean andstandard deviation
According to the instance settings in Erkut and Ingolfsson(2008) response time from a base to an emergency pointis randomly generated from a discrete Uniform distributionon interval [3 30] in units of minutes A base covers anemergency point if the driving time between them is lessthan 10 minutes to ensure the survival probability (Erkut andIngolfsson 2006) For an emergency point in each instanceif there is no base with 10 minutes driving time to it then thebase covering this point is set to be the nearest one The costof constructing a base and that of locating one ambulance areboth set to be 1 as in Nickel et al [8] Besides the number ofemergency points is 3 times that of potential base locations
532 Results and Discussion Different values of requiredservice level 120572 are tested ie 085 09 095 Computationalresults on 120 randomly generated instances are reported inTables 4-6
Table 4 reports the numerical results of instances withdemand generated from Uniform distribution We observethat when 120572 = 085 09 095 the average computationaltimes of the sampling approach are 4482 4379 and 4295seconds respectively However both approximated MIP for-mulations can produce solutions in much faster speeds iewithin 13 seconds on average With the increase of 120572 (i)the out-of-sample average service level is getting higher and(ii) the system costs of solutions obtained by all method aregetting larger and (iii) the system costs of solutions yieldedbyMIP-DF2 is getting closer to those of solutions obtained bythe sampling approach In terms of the average system costthe sampling approach performs the best while MIP-DF1 isthe poorest When 120572 = 085 for example the average systemcost ofMIP-DF2 and the sampling approach are about 3032and 442 less than that of MIP-DF1 respectively When120572 = 095 the above gaps are even larger Concluding fortheUniformdistribution (1) the proposed two approximatedMIP formulations are far more time saving than the sampling
approach (2)MIP-DF1 andMIP-DF2 can improve the servicelevel by about 1907 compared with the sampling approachhowever (3) despite of the highest service levels obtainedby MIP-DF1 the system cost obtained by MIP-DF1 is alsohigh (4) MIP-DF2 can ensure an appropriate service levelwith much less cost than MIP-DF1 and (5) for the situationswith small-scale instances and high service level requirementMIP-DF1 is recommended
Computational results of instances with demand at eachemergency point generated from Poisson distribution arepresented in Table 5 For MIP-DF1 and MIP-DF2 onlymean and variance are involved Therefore with givenmean and standard deviation of the demand at each pointcomputational results are the same for different probabilitydistributions We can obtain from Table 5 that the objectivevalue obtained by the sampling approach increases with theincrease of 120572 Besides when the value of 120572 equals 085the average service level obtained by the sampling approachis about 8143 which is smaller than that of the twoapproximated MIP formulations The highest service level9934 is obtained by MIP-DF1 and that obtained by MIP-DF2 is 9708 When 120572 = 09 the average service levelof the sampling approach is 8437 and those obtained byMIP-DF1 and MIP-DF2 are 9985 and 9797 respectivelyWhen 120572 = 095 the average service level obtained by thesampling approach is 9356 and those yielded by MIP-DF1 and MIP-DF2 are 9997 and 9781 respectively ForPoisson distribution concluding MIP-DF1 and MIP-DF2can improve the service level compared with the samplingapproach especially MIP-DF1 can obtain the highest servicelevel
Similarly Table 6 reports the computational results forinstances in which the demand at each emergency point isgenerated following Normal distribution When 120572 = 085the average service levels obtained by the sampling approachMIP-DF1 and MIP-DF2 are 8174 9925 and 9742respectively When 120572 = 09 the three values are 84579981 and 9743 respectively Moreover for 120572 = 095 thevalues are equal to 8732 9999 and 9794 respectivelyIt can be observed that the performance of the samplingapproach is the poorest in terms of the service level
By observing the above computational results on bench-marks and randomly generated instances we conclude that(1)The proposed approximated MIP formulations ie MIP-DF1 and MIP-DF2 are very time saving compared with thesampling approach (2) MIP-DF1 an MIP-DF2 can achievethe requirement of service level and improve the service levelby about 1547 and 1229 on average compared with thatof the sampling approach (3) MIP-DF1 is more conservativein terms of overhigh emergency service level close to 100and the system cost is also high which is about 568 largerthan that of MIP-DF2 (4) Compared with MIP-DF1 MIP-DF2 is less conservative and much more cost saving whileensuring an appropriately service level on average which isabout 97
In sum for medium and large problem instances (egShanghai) which are common in practice to guarantee therequired service level with less cost we recommendMIP-DF2as solution method However for very small instances with
8 Journal of Advanced Transportation
Table4Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mun
iform
distrib
ution
Set (|119868||119869|)
120572=085
120572=09
120572=095
Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
1(26)
27
897317
01
149
9906
01
107
9222
28
967373
01
155
9933
01
107
9167
27
106
7585
01
185
9987
01
107
9128
2(412
)47
169
7423
00
291
9919
00
207
9122
58
181
7438
01
303
9997
01
207
9363
50
199
7702
00
363
9982
00
207
9338
3(618
)72
258
7506
01
438
9976
01
312
9106
75279
8082
01
456
9996
01
312
9369
78304
8337
01
546
9997
01
312
9357
4(824)
99314
7956
01
553
9993
01
385
9206
111
338
8202
01
577
9999
01
385
9602
112
366
8492
01
697
9994
01
385
9620
5(1030)
132
415
8406
01
722
9896
01
512
9406
143
441
8322
02
752
9983
01
512
9692
142
476
8609
01
902
9993
01
512
9692
6(1236)
170
475
8249
02
843
9870
01
591
9367
181
504
8251
01
879
9982
01
591
9608
189
535
8726
01
1059
9997
01
591
9609
7(14
42)
215
573
8111
02
1006
9849
02
712
9606
243
609
8217
02
1048
9953
02
712
9589
233
644
8608
02
1258
9992
02
712
9582
8(1648)
248
686
7532
02
1185
9896
02
849
9725
284
727
8307
02
1233
9949
02
849
9613
281
768
9063
02
1473
9999
03
849
9622
9(1854)
313
691
8353
03
1236
9974
04
858
9709
351
731
8397
03
1290
9988
03
858
9765
352
760
8906
03
1560
9994
03
858
9758
10(2060)
372
794
8174
03
1409
9887
03
989
9657
428
843
8489
03
1469
9947
03
989
9773
432
892
9004
03
1769
9997
03
989
9738
11(2266
)451
830
8072
04
1493
9949
03
1031
9758
518
880
8328
04
1559
9982
03
1031
9723
516
930
8967
05
1889
9995
03
1031
9727
12(2472)
543
1025
7970
04
1771
9927
04
1267
9835
601
1083
8482
05
1843
9949
04
1267
9776
637
1146
9014
04
2203
9993
04
1267
9777
13(2678)
631
1026
8142
05
1819
9898
04
1274
9782
743
1087
8678
05
1897
9936
04
1274
9778
765
1146
9023
04
2287
9997
05
1273
9837
14(2884)
782
1101
7903
05
1949
9869
08
1361
9804
878
1165
8790
05
2033
9935
07
1361
9889
870
1227
9056
05
2453
9998
06
1361
9808
15(3090)
907
1091
8098
06
1997
9915
06
1367
9629
1018
1155
8329
06
2087
9981
06
1367
9807
1020
1225
8986
06
2537
9999
08
1367
9817
16(3296)
1062
1160
8798
07
2123
9944
07
1449
9787
1183
1229
8383
07
2219
9988
06
1449
9849
1193
1300
9105
06
2698
9998
06
1449
9815
17(34102)
1202
1298
7572
07
2328
9969
07
1614
9821
1377
1374
8860
08
2430
9997
07
1614
9824
1260
1452
9085
07
2940
9999
08
1614
9818
18(36108)
1394
1408
8121
08
2509
9947
08
1752
9809
1544
1492
8928
08
2617
9983
08
1752
9799
1548
1580
9173
08
3156
9999
08
1752
9809
19(38114
)1618
1508
8170
09
2676
9899
111877
9742
1760
1598
9130
09
2790
9934
08
1877
9908
1755
1688
9186
09
3360
9999
09
1877
9872
20(40120)
2206
1628
7719
102869
9989
102029
9873
2007
1732
8753
112990
9997
09
2029
9879
1904
1827
9138
103590
9998
09
2029
9875
21(42126)
2780
1622
8354
102900
9978
102017
9732
2609
1718
8854
103026
9876
102017
9818
2670
1812
9089
123655
9999
102017
9848
22(44132)
3189
1735
8494
113088
9985
152165
9857
2970
1840
8922
123220
9925
122165
9886
2873
1952
9206
113880
9999
112165
9850
23(46138)
3630
1937
8141
123372
9893
132406
9796
3222
2048
8937
123510
9934
142406
9906
3461
2170
8989
124199
10000
122406
9809
24(48144)
4226
1968
8226
133461
9932
132453
9814
3674
2085
8846
133605
9956
152453
9912
3556
2211
9300
134326
9999
132453
9819
25(50150)
4727
1919
8111
143456
9999
152407
9629
4019
2037
8699
143606
9999
142407
9857
4257
2161
9259
154356
9995
142407
9819
26(52156)
5300
1906
8096
153484
9878
172392
9658
4495
2021
8952
143640
9999
152392
9568
4486
2142
9182
1544
209999
152392
9854
27(54162)
5913
2096
8387
163751
9846
172617
9761
5117
2221
9045
163913
9998
172617
9544
4830
2355
9346
184723
9999
162617
9893
28(56168)
6310
2110
7758
173819
9933
192643
9829
5777
2234
9157
183987
9998
182643
9526
5550
2365
9045
184827
9999
172643
9831
29(58174)
6780
2175
7998
183937
9898
182720
9714
6303
2307
9022
20
4111
9969
192720
9623
6189
2444
9128
194981
10000
192720
9914
30(60180)
7593
2384
8321
194240
9967
20
2979
9841
6791
2527
8978
1944
199979
23
2979
9721
6878
2679
9143
21
5316
9999
20
2979
9869
31(62186)
8220
2281
7604
20
4157
9984
21
2856
9796
7734
2416
8934
21
4343
9997
21
2856
9679
7466
2557
9158
22
5273
9999
23
2856
9902
32(64192)
8766
2517
8378
21
4489
9932
25
3145
9729
8276
2671
8296
23
4680
9977
24
3145
9825
7897
2822
9089
24
5640
9999
22
3145
9891
33(66198)
10123
2532
7806
23
4555
9967
23
3169
9615
9010
2682
8947
24
4752
9989
24
3169
9703
9051
2835
9157
24
5742
9998
24
3169
9867
34(68204)
11010
2535
7934
24
4602
9999
29
3174
9631
9898
2685
8858
30
4806
9998
28
3174
9762
9863
2844
9236
26
5824
9999
25
3174
9889
35(70210)
11666
2714
8062
25
4862
9986
29
3391
9787
1066
62875
9137
26
5072
9999
29
3391
9716
10822
3043
9315
26
6122
9999
29
3391
9923
36(72216)
12011
2827
8104
26
5047
9991
28
3535
9798
1246
52984
9016
28
5263
9999
28
3535
9878
11667
3169
9253
30
6343
9999
29
3535
9896
37(74222)
11437
2826
8228
28
5096
9933
29
3543
9884
13582
3098
8895
29
5318
9999
32
3543
9888
12654
3178
9139
29
6427
10000
31
3543
9893
38(76228)
12435
2839
8152
29
5161
9897
29
3565
9788
13976
3103
8926
29
5388
9999
31
3565
9724
13837
3185
9255
33
6528
10000
31
3565
9895
39(78234)
1464
32945
8412
31
5331
9998
32
3695
9876
14794
3210
9005
31
5565
9999
34
3695
9713
14561
3313
9081
34
6735
10000
33
3695
9878
40(80240)
1604
22975
8510
32
5414
9957
36
3734
9796
16263
3346
8999
35
5654
9999
35
3734
9806
15869
3343
9248
35
6854
10000
35
3734
9873
Average
4482
1584
680
67
1228
397
9936
131978
796
82
4379
16913
8654
1329
626
9975
131978
79721
4295
1778
889
85
1335774
9997
131978
79775
Journal of Advanced Transportation 9
Table5Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mPo
isson
distr
ibution
set ( |119868 |
|119869 |)120572=
085120572=
09120572=
095Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
1(26)
27
907294
9967
9001
27
987770
9979
9013
24
103
9273
9999
9468
2(412
)46
170
7961
9921
9117
51
183
8305
9946
9128
49
201
9392
9998
9546
3(618
)68
258
7578
9938
8911
74280
8028
9968
9004
69
307
9182
9999
9658
4(824)
93315
8001
9928
9196
102
338
8438
9967
9172
92368
9468
9999
9613
5(1030)
120
413
8041
9939
9449
133
443
8447
9976
9272
121
479
9544
9999
9752
6(1236)
155
476
8084
9877
9554
167
508
8863
9991
9569
159
536
9456
9999
9734
7(14
42)
195
573
7876
9892
9663
213
609
9124
9998
9729
205
643
9314
10000
9768
8(1648)
232
687
7964
9997
9788
267
724
8523
9988
9687
238
772
9422
9998
9689
9(1854)
293
692
8267
9893
9809
327
730
8386
9999
9750
298
757
9228
9999
9755
10(2060)
356
796
7946
9899
9801
399
840
8299
9976
9686
360
885
9668
9999
9705
11(2266)
423
829
8201
9895
9756
489
877
8378
9987
9788
448
927
9476
9999
9812
12(2472)
512
1023
8338
9913
9853
554
1085
8501
9999
9881
514
1148
9398
9999
9888
13(2678)
592
1045
8416
9907
9873
687
1088
7998
9996
9802
607
1152
9504
9999
9821
14(2884)
723
1099
7946
9915
9788
838
1166
8056
9992
9829
776
1230
9546
9999
9839
15(3090)
855
1089
7915
9964
9703
942
1156
8432
9984
9778
903
1225
9223
9998
9785
16(3296)
1009
1158
8024
9897
9769
1163
1231
8500
9996
9824
1045
1222
9199
9999
9835
17(34102)
1165
1299
8034
9919
9779
1285
1375
8495
9937
9808
1190
1457
9258
10000
9846
18(36108)
1254
1411
8176
9943
9885
1479
1492
8561
9968
9868
1345
1579
9286
9999
9875
19(38114
)1504
1506
7970
9923
9842
1651
1594
8167
9972
9846
1564
1694
9376
9999
9786
20(40120)
1706
1635
7819
9916
9783
1854
1734
8607
9968
9776
1778
1825
9285
9904
9718
21(42126)
2054
1622
8354
9918
9834
2357
1720
8297
9978
9833
2226
1856
9501
9999
9834
22(44132)
2400
1736
8494
9922
9846
2852
1844
8269
9999
9816
2680
1946
9486
9999
9820
23(46138)
2651
1934
8141
9953
9793
3159
2049
8492
9969
9799
3072
2172
9358
10000
9813
24(48144)
3004
1961
8226
9947
9681
3522
2086
8513
9998
9706
3102
2209
9488
9999
9724
25(50150)
3308
1924
8111
9965
9746
3898
2038
8264
9999
9788
3639
2161
9546
9999
9813
26(52156)
3669
1911
8096
9939
9855
4305
2027
8722
9967
9874
4453
2142
9185
9999
9876
27(54162)
4418
2095
8387
9927
9754
4771
2219
8341
9989
9749
4806
2353
9289
9999
9755
28(56168)
5367
2113
8458
9968
9848
5356
2237
8452
9978
9855
5221
2369
9228
10000
9868
29(58174)
6217
2186
7998
9989
9694
5942
2315
8316
9996
9699
5853
2371
9167
9999
9705
30(60180)
7263
2382
8321
9991
9837
6527
2543
8359
9998
9886
6214
2678
9369
9997
9886
31(62186)
8214
2346
8404
9936
9888
7813
2430
8321
9999
9846
6876
2560
9358
9999
9847
32(64192)
8863
2518
8378
9938
9807
8106
2677
8356
9999
9833
7877
2822
9408
10000
9835
33(66198)
9985
2532
8406
9917
9709
9684
2689
8497
9996
9746
8088
2838
9507
9999
9749
34(68204)
11264
2534
8543
9908
9897
10986
2685
8809
9996
9782
8469
2843
8844
9998
9780
35(70210)
12035
2715
8258
9937
9799
12098
2886
8569
9991
9809
9946
3041
9459
9999
9813
36(72216)
12803
2825
8104
9929
9827
13076
3010
8651
9994
9834
11013
3170
9468
9999
9865
37(74222)
13021
2830
8228
9949
9826
13261
3009
8376
9999
9826
12014
3172
9346
9999
9888
38(76228)
13862
2843
8152
9933
9807
13995
3028
8760
9998
9841
12133
3187
9706
9999
9876
39(78234)
15387
2953
8312
9964
9799
14827
3133
9001
9995
9783
13381
3304
8913
9999
9799
40(80240)
16974
2975
8510
9985
9767
15256
3249
8217
9993
9781
15867
3347
9111
9999
9791
Average
4352
15875
8143
9934
9708
4362
16856
8437
9985
9707
3968
17763
9356
9997
9781
10 Journal of Advanced Transportation
Table6Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mno
rmaldistr
ibution
set ( |119868 |
|119869 |)120572=
085120572=
09120572=
095Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
1(26)
27
907294
9964
9117
27
947844
9970
9202
2558
106
8422
10000
9208
2(412
)46
170
7961
9921
9283
48
183
8328
9982
9283
4508
183
8823
10000
9370
3(618
)68
258
7578
9916
9228
68
280
7798
9971
9417
6803
307
7658
9998
9447
4(824)
93315
8000
9899
9367
96337
8379
9979
9467
9373
368
8438
9999
9579
5(1030)
120
413
7841
9927
9437
124
443
8763
9963
9476
1262
499
8850
9998
9708
6(1236)
155
476
7921
9919
9519
162
504
8124
9982
9582
16021
536
8954
10000
9756
7(14
42)
195
573
7976
9933
9598
196
609
8287
9975
9619
2011
646
8467
9999
9767
8(1648)
232
687
8031
9918
9705
235
727
8115
9965
9650
2404
772
8278
9999
9765
9(1854)
293
692
8086
9898
9759
291
734
8464
9974
9755
29815
796
8617
9998
9835
10(2060)
356
796
7984
9937
9770
358
842
8223
9988
9765
37973
921
8846
10000
9838
11(2266
)421
829
8101
9940
9755
436
877
8211
9986
9786
4514
8959
8497
9999
9789
12(2472)
514
1023
8033
9951
9785
501
1085
8162
9983
9788
52073
1147
8321
9999
9824
13(2678)
592
1023
8138
9916
9885
599
1087
8411
9970
9845
62937
1194
8609
9998
9795
14(2884)
728
1099
8235
9913
9915
722
1168
8435
9987
9806
75053
1270
8567
9997
9834
15(3090)
855
1089
8187
9922
9861
880
1157
8376
9976
9796
89222
1300
8611
9996
9798
16(3296)
1009
1158
8770
9897
9788
966
1231
8845
9984
9823
106288
1359
8997
9999
9848
17(34102)
1154
1299
8434
9925
9849
1137
1377
8601
9980
9798
123869
1493
8798
9996
9877
18(36108)
1407
1411
8321
9927
9866
1339
1495
8301
9965
9822
133807
1601
8567
9999
9834
19(38114
)1702
1506
8206
9915
9805
1672
1598
8977
9978
9845
152999
1694
8998
9999
9847
20(40120)
2091
1635
8528
9918
9889
1916
1737
8805
9976
9837
1890
021833
9012
9999
9791
21(42126)
2843
1622
8550
9934
9913
2432
1718
8886
9974
9813
2399
131815
9008
10000
9815
22(44132)
3300
1736
8449
9940
9859
2795
1843
8786
9977
9756
253106
1948
8997
9999
9826
23(46138)
3741
1934
8133
9911
9876
3143
2050
8876
9981
9785
274583
2173
9014
9999
9858
24(48144)
4531
1961
8245
9909
9846
3515
2090
8344
9993
9819
308089
2211
8661
9999
9883
25(50150)
4715
1924
8111
9929
9888
3959
2041
8279
9996
9837
3510
932158
8527
9999
9827
26(52156)
5223
1911
8096
9898
9857
4301
2025
8251
9997
9757
421747
2146
8497
9999
9914
27(54162)
5994
2095
8387
9908
9775
4835
2226
8497
9974
9780
4614
432359
8695
9999
9837
28(56168)
6406
2111
8411
9906
9798
5336
2240
8558
9991
9820
494113
2371
9249
9999
9816
29(58174)
6897
2176
8012
9916
9791
5948
2311
8211
9984
9765
624835
2445
8531
10000
9851
30(60180)
7436
2391
8111
9913
9766
6558
2533
8305
9968
9795
696116
2682
8575
9999
9799
31(62186)
8118
2282
8119
9928
9806
7069
2422
8417
9973
9873
797768
2564
8667
9999
9846
32(64192)
8895
2519
8223
9919
9757
7857
2675
8631
9982
9769
864813
2827
8976
10000
9905
33(66198)
11235
2533
8563
9928
9811
9162
2688
8547
9979
9828
973089
2843
8657
9999
9846
34(68204)
11587
2534
8249
9926
9847
9549
2686
8798
9938
9856
10523
3041
8999
10000
9877
35(70210)
1164
92715
8098
9895
9872
10345
2880
8602
9998
9869
11272
3052
9021
9999
9897
36(72216)
12224
2825
8403
9958
9770
11165
3002
8657
9999
9839
12456
3175
9113
9999
9864
37(74222)
12978
2830
8357
9937
9795
11850
3005
8496
9998
9894
13458
3176
8827
9999
9905
38(76228)
13426
2844
8267
9966
9759
12795
3019
8604
9999
9797
15155
3121
8978
9999
9853
39(78234)
15003
2955
8303
9956
9867
13816
3132
8588
9996
9885
15041
3310
9016
9999
9912
40(80240)
16495
2978
8257
9954
9826
14896
3160
8495
9997
9837
16563
3349
8957
9999
9933
Average
4619
15855
8174
9925
9742
4077
16828
8457
9981
9743
4350
17938
8732
9999
9794
Journal of Advanced Transportation 11
high service level requirement we recommend MIP-DF1 forsolutions
6 Conclusion
This paper investigates an ambulance location problem withambiguity set of demand in which the probability distribu-tion of the uncertain demand is unknown In such a data-driven environment only the mean and covariance demandsat emergency points are knownThe problem is to determinethe base locations and the number of ambulances at eachbase aiming at minimizing the total cost associated withlocating bases and assigning ambulances We propose adistribution-free model with chance constraints Then twoapproximated MIP formulations with different approxima-tion methods of chance constraints are proposed which arebased on different ambiguity sets of the unknown probabilitydistribution Numerical experiments on benchmarks and120 randomly generated instances are conducted and thecomputational results show that the first approximated MIPformulation is much conservative with overhigh system costwhile the second approximatedMIP formulation ismore costsaving with service level appropriately ensured
In future research wemay consider the following relevantissues First the uncertainty of driving time from a base toan emergency point shall be analysed Besides unexpected orsudden disasters may be taken into consideration Moreoverother uncertainties should be considered such as travel timethe number of available vehicles and so on The stochasticdependence should also further considered
Data Availability
The data of the benchmark instances in our manuscript canbe obtained in Nickel et al [8] For the test instances thedata is randomly generated and the way for generating data isstated in our manuscript The datasets generated during thecurrent study are available from the corresponding author onreasonable request
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This work was supported by the National Natural Sci-ence Foundation of China (NSFC) under Grants 7142800271531011 71571134 and 71771048 This work was also sup-ported by the Fundamental Research Funds for the CentralUniversities
References
[1] L Brotcorne G Laporte and F Semet ldquoAmbulance loca-tion and relocation modelsrdquo European Journal of OperationalResearch vol 147 no 3 pp 451ndash463 2003
[2] X Li Z Zhao X Zhu and T Wyatt ldquoCovering modelsand optimization techniques for emergency response facilitylocation and planning a reviewrdquo Mathematical Methods ofOperations Research vol 74 no 3 pp 281ndash310 2011
[3] C Toregas R Swain C ReVelle and L Bergman ldquoThe locationof emergency service facilitiesrdquoOperations Research vol 19 no6 pp 1363ndash1373 1971
[4] R Church and C ReVelle ldquoThe maximal covering locationproblemrdquo Papers in Regional Science vol 32 no 1 pp 101ndash1181974
[5] M Gendreau G Laporte and F Semet ldquoSolving an ambulancelocation model by tabu searchrdquo Location Science vol 5 no 2pp 75ndash88 1997
[6] A Ingolfsson S Budge and E Erkut ldquoOptimal ambulancelocation with random delays and travel timesrdquo Health CareManagement Science vol 11 no 3 pp 262ndash274 2008
[7] M Gendreau G Laporte and F Semet ldquoThemaximal expectedcoverage relocation problem for emergency vehiclesrdquo Journal ofthe Operational Research Society vol 57 no 1 pp 22ndash28 2006
[8] S Nickel M Reuter-Oppermann and F Saldanha-da-GamaldquoAmbulance location under stochastic demand A samplingapproachrdquo Operations Research for Health Care vol 8 pp 24ndash32 2016
[9] P Beraldi M E Bruni and D Conforti ldquoDesigning robustemergency medical service via stochastic programmingrdquo Euro-pean Journal of Operational Research vol 158 no 1 pp 183ndash1932004
[10] N Noyan ldquoAlternate risk measures for emergency medicalservice system designrdquo Annals of Operations Research vol 181pp 559ndash589 2010
[11] M R Wagner ldquoStochastic 0-1 linear programming underlimited distributional informationrdquoOperations Research Lettersvol 36 no 2 pp 150ndash156 2008
[12] EDelage andYYe ldquoDistributionally robust optimization undermoment uncertainty with application to data-driven problemsrdquoOperations Research vol 58 no 3 pp 595ndash612 2010
[13] J T Essen J L Hurink S Nickel and M Reuter ldquoModels forambulance planning on the strategic and the tactical levelrdquo BetaResearch School for Operations Management and Logistics p 272013
[14] E Erkut A Ingolfsson and G Erdogan ldquoAmbulance locationfor maximum survivalrdquoNaval Research Logistics (NRL) vol 55no 1 pp 42ndash58 2008
[15] S C Chapman and J A White ldquoProbabilistic formulations ofemergency service facilities location problemsrdquo in Proceedingsof ORSATIMS Conference San Juan Puerto Rico 1974
[16] A Ben-Tal D Den Hertog A De Waegenaere B Melenbergand G Rennen ldquoRobust solutions of optimization problemsaffected by uncertain probabilitiesrdquo Management Science vol59 no 2 pp 341ndash357 2013
[17] MNg ldquoDistribution-free vessel deployment for liner shippingrdquoEuropean Journal of Operational Research vol 238 no 3 pp858ndash862 2014
[18] M Ng ldquoContainer vessel fleet deployment for liner shippingwith stochastic dependencies in shipping demandrdquo Transporta-tion Research Part B Methodological vol 74 pp 79ndash87 2015
[19] R Jiang and Y Guan ldquoData-driven chance constrained stochas-tic programrdquo Mathematical Programming vol 158 no 1-2 SerA pp 291ndash327 2016
[20] C-M Lee and S-L Hsu ldquoThe effect of advertising on thedistribution-free newsboy problemrdquo International Journal ofProduction Economics vol 129 no 1 pp 217ndash224 2011
[21] K Kwon and T Cheong ldquoA minimax distribution-free proce-dure for a newsvendor problem with free shippingrdquo EuropeanJournal of Operational Research vol 232 no 1 pp 234ndash2402014
12 Journal of Advanced Transportation
[22] Y Zhang S Shen and S A Erdogan ldquoDistributionally robustappointment scheduling with moment-based ambiguity setrdquoOperations Research Letters vol 45 no 2 pp 139ndash144 2017
[23] Y Zhang S Shen and S A Erdogan ldquoSolving 0-1semidefinite programs for distributionally robust allocationof surgery blocksrdquo Social Science Electronic Publishing 2017httpsssrncomabstract=2997524
[24] F Zheng J He F Chu and M Liu ldquoA new distribution-freemodel for disassembly line balancing problem with stochastictask processing timesrdquo International Journal of ProductionResearch pp 1ndash13 2018
[25] J Cheng E Delage and A Lisser ldquoDistributionally robuststochastic knapsack problemrdquo SIAM Journal on Optimizationvol 24 no 3 pp 1485ndash1506 2014
[26] M Liu D Yang and F Hao ldquoOptimization for the locationsof ambulances under two-stage life rescue in the emergencymedical service a case study in Shanghai ChinardquoMathematicalProblems in Engineering vol 2017 Article ID 1830480 14 pages2017
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
4 Journal of Advanced Transportation
The objective function denotes the goal to minimize thetotal cost consisting of two parts (i) the cost for constructingbases ie sum119894isin119868 119891119894 sdot 119909119894 and (ii) the cost for deploying ambu-lances ie sum119894isin119868 119892119894 sdot 119911119894
Constraint (3) ensures that ambulances can only belocated at the opened bases Constraint (4) ensures that thenumber of ambulances sent to serve emergency points frombase 119894 isin 119868 does not exceed the total number of ambulanceslocated at 119894 isin 119868 Constraints (5)-(7) are the restrictions ondecision variables
4 Solution Approaches
The proposed distribution-free model is difficult to solvewith the commercial software due to chance constraints Inthis section we propose two approximatedMIP formulationsbased on those in Wagner [11] and in Delage and Ye [12]respectively In the following subsections we present the twoapproximated MIP formulations
41 Approximated MIP Formulation MIP-DF1 In this partwe first describe a widely used ambiguity set to describe theuncertainty Then an approximated MIP formulation MIP-DF1 is proposed
Given a set of independent historical data samples 119889119904|119878|119904=1of random vectors of demands where 119878 is the set of sampleindexes the mean vector 120583 and the covariance matrix Σ ofdemands can be estimated as follows
120583 = 1|119878|sum119904isin119878119889119904Σ = 1|119878|sum119904isin119878 (119889119904 minus 120583) (119889119904 minus 120583)⊤
(8)
where (sdot)⊤ implies the transposition of the vector in paren-theses Then an ambiguity set of all probability distributionsof demandsP1(120583 Σ) can be described as the following
P1 (120583 Σ) = P EP [119889] = 120583EP [(119889 minus 120583) (119889 minus 120583)⊤] = Σ (9)
whereP denotes a possible probability distribution satisfyingthe given conditions and EP[sdot] denotes the expected valueof the number in parentheses Then the chance constraint (1)with the ambiguity setP1 can be presented as follows
ProbP (sum119894isin119868119895
119910119894119895 ge 119889119895) ge 120572119895 forall119895 isin 119869 P isin P1 (10)
According to Wagner (2010) and Ng [18] constraint (1) canbe conservatively approximated by the following
sum119894isin119868119895
119910119894119895 ge 120583119895 + 120590119895 sdot radic 120572119895(1 minus 120572119895) forall119895 isin 119869 (11)
where 120590119895 = radicΣ119895119895 and Σ119895119895 denotes the 119895-th element in the 119895-thcolumn of matrix Σ In terms of conservative approximation
all solutions satisfying constraint (11) must satisfy constraint(1) Then we describe the approximated MIP formulationMIP-DF1 toDF combined with constraint (11)
[MIP-DF1]
min sum119894isin119868
(119891119894 sdot 119909119894 + 119892119894 sdot 119911119894)st Constraints (4) - (8) (12)
(12)
MIP-DF1 can be exactly solved by the commercialsoftware such as CPLEX As we can observe from thecomputational results in Section 5 it can obtain a relativelyhigh service level However the system cost is also high Inorder to save the system cost another approximated MIPformulation of DF is then proposed in the next subsection
42 Approximated MIP Formulation MIP-DF2 AmbiguitysetP1 focuses on exactly matching the mean and covariancematrix of uncertain parameters However in practice theremay be considerable estimation errors in the mean andcovariance matrix To take the inevitable estimation errorsinto consideration Delage and Ye [12] introduce a newmoment-based ambiguity set Therefore we in the followingemploy the moment-based ambiguity set
P2 (120583 Σ 1205741 1205742)= P (EP [119889] minus 120583)⊤ Σminus1 (EP [119889] minus 120583) le 1205741
EP [(119889 minus 120583) (119889 minus 120583)⊤] ⪯ 1205742Σ (13)
where 1205741 ge 0 and 1205742 ge 0 are two parameters of ambiguityset P2(120583 Σ 1205741 1205742) with the following assumptions (i) thetrue mean vector of demands is within an ellipsoid of sizeproportion to 1205741 centered at 120583 and (ii) the true covariancematrix of demands is in a positive semidefinite cone boundedby amatrix inequality of 1205742ΣThe ambiguity set describes howlikely the uncertain parameters are to be close to the meanin terms of the correlation [12] Besides under the moment-based ambiguity set various stochastic programs with partialdistributional information have been successfully modeledand solved [22 23 25]
According to the approximation method in Zhang et al[23] chance constraint (1) can be approximated by
radic 11 minus 119886 minus 119887 sdot (1 + radic1 minus 120572119895120572119895 sdot 119887) sdot 120590119895le radic1 minus 120572119895120572119895 sdot (sum
119894isin119868119895
119910119894119895 minus 120583119895) 119895 isin 119869(14)
where
119886 = 1 minus 1205741 + 11205742 minus 1205741 119887 = 12057411205742 minus 1205741
(15)
Journal of Advanced Transportation 5
Table 1 120572 = 095|119868| = |119869| Sampling approach MIP-DF1 MIP-DF2
CT Obj Sel () CT Obj Sel () CT Obj Sel ()1 7 49 23 9626 01 36 10000 01 23 100002 7 51 22 9863 01 35 10000 01 24 99823 6 41 14 9692 01 30 10000 01 14 98504 6 43 13 9713 02 29 10000 01 12 97425 6 57 12 7996 01 30 10000 02 12 100006 8 58 21 9850 01 24 10000 02 27 10000
Average 50 175 9457 01 307 10000 01 187 9929
The second approximated MIP formulationMIP-DF2 toDF is presented as follows
[MIP-DF2]
min sum119894isin119868
(119891119894 sdot 119909119894 + 119892119894 sdot 119911119894)st Constraints (4) - (8) (15)
(16)
Notice that when 1205741 = 0 and 1205742 = 1 inequality (14)reduces to (11) in MIP-DF1 which implies that MIP-DF1 isa special case of MIP-DF2
5 Computational Experiments
In this section the performance of our proposed formula-tions first evaluated and compared to the sampling approachproposed in Nickel et al [8] Specifically given a requiredservice level 120572 we apply our MIP formulations by restrictingthe chance constraint for each emergency point to satisfya safety level 120572119895 = 120572 while we adopt the samplingapproach by restricting the coverage constraint with 120572 Thencomputational results on 120 randomly generated instancesbased on the information on emergencies in Shanghai arereported Our approximated MIP formulations and thesampling approach are solved by coding in MATLAB 2014bcalling CPLEX 126 solver All numerical experiments areconducted on a personal computer with Core I5 and 330GHzprocessor and 8GB RAMunder windows 7 operating systemThe computational time for sampling approach is limited to3600 seconds
51 Out-of-Sample Test FollowingZhang et alrsquos [22]method-ology we examine the out-of-sample performance of solu-tions obtained by our proposed approximated MIP formula-tions and the sampling approach [8] The out-of-sample testincludes three steps
(1) The base locations and the employment of ambu-lances at each base are determined with known meanvector and covariance matrix of demands which isnamed as in-sample test
(2) 1000 scenarios are sampled from the underlying truedistribution which represent realizing the uncertaindemand
(3) Based on the solution obtained in step (1) ambu-lances are assigned to serve emergency points inthe 1000 scenarios to maximize the (out-of-sample)service level
The (out-of-sample) service level is defined as the ratioof the number of emergency points with demand satisfiedto the total number of emergency points in 1000 scenariosIn the following subsections the benchmarks are tested andthen three common distributions are employed ie uniformdistribution Poisson distribution and normal distributionrespectively
52 Benchmark Instances We first test the benchmarks inNickel et al [8] to compare our proposed two approximatedMIP formulationswith the sampling approach InNickel et al[8] they assume that demand nodes are also potential baselocations and the number of ambulances required by eachemergency point is chosen from set 0 1 2with given proba-bility distribution Computational results on benchmarks arereported in Tables 1 and 2
Table 1 reports the computational results with servicelevel requirement120572 = 095 In Table 1 CTObj and Sel denotethe computational time objective value (ie the system cost)and the service level respectively The first column includesthe indexes of instances and the second indicates the numberof nodes It shows that the average service level obtainedby the sampling approach is 9457 which is smaller thanthe required service level ie 85 Specifically the servicelevel of instance 5 obtained by the sampling approach is only7996 However MIP-DF1 and MIP-DF2 can guarantee theservice levels for all instances higher than the requirementConcluding from Table 1 we can observe that (i) the averagecomputational time of sampling approach is about 50 timesas those of the two approximated MIP formulations and (ii)MIP-DF1 and MIP-DF2 improve the average service level byabout 574 and 499 compared with sampling approachhowever (iii) system costs obtained by MIP-DF1 and MIP-DF2 are respectively 7543 and 686 higher than that ofsampling approach
Table 2 reports the computational results with servicelevel requirement 120572 = 099 For instances 2-6 the ser-vice levels obtained by sampling approach cannot meet therequirement ie 99 Besides it shows that the averagecomputational time of sampling approach is 48 which isabout 48 times greater than those of MIP-DF1 andMIP-DF2
6 Journal of Advanced Transportation
Table 2 120572 = 099|119868| = |119869| Sampling approach MIP-DF1 MIP-DF2
CT Obj Sel () CT Obj Sel () CT Obj Sel ()1 7 49 24 9979 01 64 10000 01 25 100002 7 48 23 9856 01 63 10000 01 24 100003 6 51 15 9717 01 55 10000 01 16 100004 6 47 15 9275 01 52 10000 01 13 100005 6 40 14 9000 01 55 10000 01 14 100006 8 52 23 9785 01 45 10000 01 27 10000
Average 48 190 9343 01 557 10000 01 198 10000
Table 3 Solutions to instance 4 in Nickel et al [8] for the given scenario
Sampling approach MIP-DF1 MIP-DF2Emergency points 1 2 3 4 5 6 Emergency points 1 2 3 4 5 6 Emergency points 1 2 3 4 5 6
Demand 1 0 2 0 2 0 Demand 1 0 2 0 2 0 Demand 1 0 2 0 2 0Base locations Ambulance employment Ambulance employment Ambulance employment1 0 0 02 1 1 0 3 1 23 1 1 26 1 2 2 04 0 0 05 0 6 06 2 2 0 4 2
The average service level is 9343 Although the averageservice levels obtained by MIP-DF1 and MIP-DF2 are both100 the average system cost obtained by MIP-DF1 is about3 times larger than that of sampling approach Besides theaverage system cost obtained by MIP-DF2 is 421 higherthan that of sampling approach
Concluding for the benchmarks described in Nickel etal [8] we can observe that (1) MIP-DF1 and MIP-DF2 canbe solved in far less computational time compared with thesampling approach (2) the sampling approach cannot meetthe requirement of service level for all benchmarks whilethe two MIP formulations can (3) the average system cost ofMIP-DF2 is very close to that of sampling approach (4) withthe increase of 120572 the average system costs obtained by MIP-DF2 is getting closer to that obtained by sampling approach
From Tables 1 and 2 we can observe that in spite ofthe lower system cost obtained by the sampling approachthere may exist extreme situations that demands of someemergency points are not satisfied Take one benchmark inNickel et al [8] ie instance 4 where there are 6 nodesA scenario in which demand at each emergency point isgenerated from the given probability distribution is shown inFigure 1 We can observe from Figure 1 that under the givenscenario the demand of each possible emergency point is1 0 2 0 2 0The solutions obtained by sampling approachMIP-DF1 andMIP-DF2 are shown in Table 3 which includesbase locations and ambulance employment obtained by in-sample test and ambulance assignment under given scenario
In Table 3 the first column includes the indexes of poten-tial base locations The second column represents the baselocations and the ambulance employment at each location
>1 = 1
>2 = 0
>3 = 2
>4 = 0
>5 = 2
>6 = 0
1
2 4
3 5
6
Figure 1 Instance 4 in Nickel et al [8] under given scenario
in which a location with 0 ambulance implies that it is notselected for base construction From Table 3 we can observethat the in-sample solutions obtained by sampling approachis to construct bases at locations 2 3 and 6 where the numberof ambulances is 1 1 and 2 respectively Besides undergiven scenario there is one ambulance serving emergencypoint 1 from base location 3 and one ambulance servingemergency point from base location 2 and 2 ambulancesserving emergency point 5 from base location 6 It shows thatthe requirement of ambulances at emergency point 3 is notfully satisfied The insufficient service may cause serious lossof life However we can find that the solutions obtained byMIP-DF1 andMIP-DF2 can provide sufficient ambulances forall emergency points
In sum to guarantee the required service level withless cost MIP-DF2 is recommended for the solutions with
Journal of Advanced Transportation 7
Nickel et al [8]rsquos data set Moreover in the benchmarks itis assumed that all nodes are possible emergency points aswell as potential base locations and the number of requiredambulances is small which is not always consistent withthe situation in Shanghai Therefore we further test the twoapproximated MIP formulations and the sampling approachon randomly generated instances by analysing the informa-tion on emergencies in Shanghai [26]
53 Computational Tests on Randomly Generated InstancesIn the following subsection numerical experiments on 120randomly generated instances are described
531 Test Instances By analysing the information on emer-gencies in Shanghai [26] the mean value 120583119895 of demand 119889119895 ineach emergency point are randomly generated fromadiscreteUniform distribution on interval [5 25] The standard devia-tion 120590119895 is set to be 5 Since the sampling approach proposed inNickel et al [8] requires a sample of scenarios in numericalexperiments scenarios are produced with demands randomlygenerated from Uniform distribution Poisson distributionand Normal distribution consistent with given mean andstandard deviation
According to the instance settings in Erkut and Ingolfsson(2008) response time from a base to an emergency pointis randomly generated from a discrete Uniform distributionon interval [3 30] in units of minutes A base covers anemergency point if the driving time between them is lessthan 10 minutes to ensure the survival probability (Erkut andIngolfsson 2006) For an emergency point in each instanceif there is no base with 10 minutes driving time to it then thebase covering this point is set to be the nearest one The costof constructing a base and that of locating one ambulance areboth set to be 1 as in Nickel et al [8] Besides the number ofemergency points is 3 times that of potential base locations
532 Results and Discussion Different values of requiredservice level 120572 are tested ie 085 09 095 Computationalresults on 120 randomly generated instances are reported inTables 4-6
Table 4 reports the numerical results of instances withdemand generated from Uniform distribution We observethat when 120572 = 085 09 095 the average computationaltimes of the sampling approach are 4482 4379 and 4295seconds respectively However both approximated MIP for-mulations can produce solutions in much faster speeds iewithin 13 seconds on average With the increase of 120572 (i)the out-of-sample average service level is getting higher and(ii) the system costs of solutions obtained by all method aregetting larger and (iii) the system costs of solutions yieldedbyMIP-DF2 is getting closer to those of solutions obtained bythe sampling approach In terms of the average system costthe sampling approach performs the best while MIP-DF1 isthe poorest When 120572 = 085 for example the average systemcost ofMIP-DF2 and the sampling approach are about 3032and 442 less than that of MIP-DF1 respectively When120572 = 095 the above gaps are even larger Concluding fortheUniformdistribution (1) the proposed two approximatedMIP formulations are far more time saving than the sampling
approach (2)MIP-DF1 andMIP-DF2 can improve the servicelevel by about 1907 compared with the sampling approachhowever (3) despite of the highest service levels obtainedby MIP-DF1 the system cost obtained by MIP-DF1 is alsohigh (4) MIP-DF2 can ensure an appropriate service levelwith much less cost than MIP-DF1 and (5) for the situationswith small-scale instances and high service level requirementMIP-DF1 is recommended
Computational results of instances with demand at eachemergency point generated from Poisson distribution arepresented in Table 5 For MIP-DF1 and MIP-DF2 onlymean and variance are involved Therefore with givenmean and standard deviation of the demand at each pointcomputational results are the same for different probabilitydistributions We can obtain from Table 5 that the objectivevalue obtained by the sampling approach increases with theincrease of 120572 Besides when the value of 120572 equals 085the average service level obtained by the sampling approachis about 8143 which is smaller than that of the twoapproximated MIP formulations The highest service level9934 is obtained by MIP-DF1 and that obtained by MIP-DF2 is 9708 When 120572 = 09 the average service levelof the sampling approach is 8437 and those obtained byMIP-DF1 and MIP-DF2 are 9985 and 9797 respectivelyWhen 120572 = 095 the average service level obtained by thesampling approach is 9356 and those yielded by MIP-DF1 and MIP-DF2 are 9997 and 9781 respectively ForPoisson distribution concluding MIP-DF1 and MIP-DF2can improve the service level compared with the samplingapproach especially MIP-DF1 can obtain the highest servicelevel
Similarly Table 6 reports the computational results forinstances in which the demand at each emergency point isgenerated following Normal distribution When 120572 = 085the average service levels obtained by the sampling approachMIP-DF1 and MIP-DF2 are 8174 9925 and 9742respectively When 120572 = 09 the three values are 84579981 and 9743 respectively Moreover for 120572 = 095 thevalues are equal to 8732 9999 and 9794 respectivelyIt can be observed that the performance of the samplingapproach is the poorest in terms of the service level
By observing the above computational results on bench-marks and randomly generated instances we conclude that(1)The proposed approximated MIP formulations ie MIP-DF1 and MIP-DF2 are very time saving compared with thesampling approach (2) MIP-DF1 an MIP-DF2 can achievethe requirement of service level and improve the service levelby about 1547 and 1229 on average compared with thatof the sampling approach (3) MIP-DF1 is more conservativein terms of overhigh emergency service level close to 100and the system cost is also high which is about 568 largerthan that of MIP-DF2 (4) Compared with MIP-DF1 MIP-DF2 is less conservative and much more cost saving whileensuring an appropriately service level on average which isabout 97
In sum for medium and large problem instances (egShanghai) which are common in practice to guarantee therequired service level with less cost we recommendMIP-DF2as solution method However for very small instances with
8 Journal of Advanced Transportation
Table4Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mun
iform
distrib
ution
Set (|119868||119869|)
120572=085
120572=09
120572=095
Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
1(26)
27
897317
01
149
9906
01
107
9222
28
967373
01
155
9933
01
107
9167
27
106
7585
01
185
9987
01
107
9128
2(412
)47
169
7423
00
291
9919
00
207
9122
58
181
7438
01
303
9997
01
207
9363
50
199
7702
00
363
9982
00
207
9338
3(618
)72
258
7506
01
438
9976
01
312
9106
75279
8082
01
456
9996
01
312
9369
78304
8337
01
546
9997
01
312
9357
4(824)
99314
7956
01
553
9993
01
385
9206
111
338
8202
01
577
9999
01
385
9602
112
366
8492
01
697
9994
01
385
9620
5(1030)
132
415
8406
01
722
9896
01
512
9406
143
441
8322
02
752
9983
01
512
9692
142
476
8609
01
902
9993
01
512
9692
6(1236)
170
475
8249
02
843
9870
01
591
9367
181
504
8251
01
879
9982
01
591
9608
189
535
8726
01
1059
9997
01
591
9609
7(14
42)
215
573
8111
02
1006
9849
02
712
9606
243
609
8217
02
1048
9953
02
712
9589
233
644
8608
02
1258
9992
02
712
9582
8(1648)
248
686
7532
02
1185
9896
02
849
9725
284
727
8307
02
1233
9949
02
849
9613
281
768
9063
02
1473
9999
03
849
9622
9(1854)
313
691
8353
03
1236
9974
04
858
9709
351
731
8397
03
1290
9988
03
858
9765
352
760
8906
03
1560
9994
03
858
9758
10(2060)
372
794
8174
03
1409
9887
03
989
9657
428
843
8489
03
1469
9947
03
989
9773
432
892
9004
03
1769
9997
03
989
9738
11(2266
)451
830
8072
04
1493
9949
03
1031
9758
518
880
8328
04
1559
9982
03
1031
9723
516
930
8967
05
1889
9995
03
1031
9727
12(2472)
543
1025
7970
04
1771
9927
04
1267
9835
601
1083
8482
05
1843
9949
04
1267
9776
637
1146
9014
04
2203
9993
04
1267
9777
13(2678)
631
1026
8142
05
1819
9898
04
1274
9782
743
1087
8678
05
1897
9936
04
1274
9778
765
1146
9023
04
2287
9997
05
1273
9837
14(2884)
782
1101
7903
05
1949
9869
08
1361
9804
878
1165
8790
05
2033
9935
07
1361
9889
870
1227
9056
05
2453
9998
06
1361
9808
15(3090)
907
1091
8098
06
1997
9915
06
1367
9629
1018
1155
8329
06
2087
9981
06
1367
9807
1020
1225
8986
06
2537
9999
08
1367
9817
16(3296)
1062
1160
8798
07
2123
9944
07
1449
9787
1183
1229
8383
07
2219
9988
06
1449
9849
1193
1300
9105
06
2698
9998
06
1449
9815
17(34102)
1202
1298
7572
07
2328
9969
07
1614
9821
1377
1374
8860
08
2430
9997
07
1614
9824
1260
1452
9085
07
2940
9999
08
1614
9818
18(36108)
1394
1408
8121
08
2509
9947
08
1752
9809
1544
1492
8928
08
2617
9983
08
1752
9799
1548
1580
9173
08
3156
9999
08
1752
9809
19(38114
)1618
1508
8170
09
2676
9899
111877
9742
1760
1598
9130
09
2790
9934
08
1877
9908
1755
1688
9186
09
3360
9999
09
1877
9872
20(40120)
2206
1628
7719
102869
9989
102029
9873
2007
1732
8753
112990
9997
09
2029
9879
1904
1827
9138
103590
9998
09
2029
9875
21(42126)
2780
1622
8354
102900
9978
102017
9732
2609
1718
8854
103026
9876
102017
9818
2670
1812
9089
123655
9999
102017
9848
22(44132)
3189
1735
8494
113088
9985
152165
9857
2970
1840
8922
123220
9925
122165
9886
2873
1952
9206
113880
9999
112165
9850
23(46138)
3630
1937
8141
123372
9893
132406
9796
3222
2048
8937
123510
9934
142406
9906
3461
2170
8989
124199
10000
122406
9809
24(48144)
4226
1968
8226
133461
9932
132453
9814
3674
2085
8846
133605
9956
152453
9912
3556
2211
9300
134326
9999
132453
9819
25(50150)
4727
1919
8111
143456
9999
152407
9629
4019
2037
8699
143606
9999
142407
9857
4257
2161
9259
154356
9995
142407
9819
26(52156)
5300
1906
8096
153484
9878
172392
9658
4495
2021
8952
143640
9999
152392
9568
4486
2142
9182
1544
209999
152392
9854
27(54162)
5913
2096
8387
163751
9846
172617
9761
5117
2221
9045
163913
9998
172617
9544
4830
2355
9346
184723
9999
162617
9893
28(56168)
6310
2110
7758
173819
9933
192643
9829
5777
2234
9157
183987
9998
182643
9526
5550
2365
9045
184827
9999
172643
9831
29(58174)
6780
2175
7998
183937
9898
182720
9714
6303
2307
9022
20
4111
9969
192720
9623
6189
2444
9128
194981
10000
192720
9914
30(60180)
7593
2384
8321
194240
9967
20
2979
9841
6791
2527
8978
1944
199979
23
2979
9721
6878
2679
9143
21
5316
9999
20
2979
9869
31(62186)
8220
2281
7604
20
4157
9984
21
2856
9796
7734
2416
8934
21
4343
9997
21
2856
9679
7466
2557
9158
22
5273
9999
23
2856
9902
32(64192)
8766
2517
8378
21
4489
9932
25
3145
9729
8276
2671
8296
23
4680
9977
24
3145
9825
7897
2822
9089
24
5640
9999
22
3145
9891
33(66198)
10123
2532
7806
23
4555
9967
23
3169
9615
9010
2682
8947
24
4752
9989
24
3169
9703
9051
2835
9157
24
5742
9998
24
3169
9867
34(68204)
11010
2535
7934
24
4602
9999
29
3174
9631
9898
2685
8858
30
4806
9998
28
3174
9762
9863
2844
9236
26
5824
9999
25
3174
9889
35(70210)
11666
2714
8062
25
4862
9986
29
3391
9787
1066
62875
9137
26
5072
9999
29
3391
9716
10822
3043
9315
26
6122
9999
29
3391
9923
36(72216)
12011
2827
8104
26
5047
9991
28
3535
9798
1246
52984
9016
28
5263
9999
28
3535
9878
11667
3169
9253
30
6343
9999
29
3535
9896
37(74222)
11437
2826
8228
28
5096
9933
29
3543
9884
13582
3098
8895
29
5318
9999
32
3543
9888
12654
3178
9139
29
6427
10000
31
3543
9893
38(76228)
12435
2839
8152
29
5161
9897
29
3565
9788
13976
3103
8926
29
5388
9999
31
3565
9724
13837
3185
9255
33
6528
10000
31
3565
9895
39(78234)
1464
32945
8412
31
5331
9998
32
3695
9876
14794
3210
9005
31
5565
9999
34
3695
9713
14561
3313
9081
34
6735
10000
33
3695
9878
40(80240)
1604
22975
8510
32
5414
9957
36
3734
9796
16263
3346
8999
35
5654
9999
35
3734
9806
15869
3343
9248
35
6854
10000
35
3734
9873
Average
4482
1584
680
67
1228
397
9936
131978
796
82
4379
16913
8654
1329
626
9975
131978
79721
4295
1778
889
85
1335774
9997
131978
79775
Journal of Advanced Transportation 9
Table5Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mPo
isson
distr
ibution
set ( |119868 |
|119869 |)120572=
085120572=
09120572=
095Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
1(26)
27
907294
9967
9001
27
987770
9979
9013
24
103
9273
9999
9468
2(412
)46
170
7961
9921
9117
51
183
8305
9946
9128
49
201
9392
9998
9546
3(618
)68
258
7578
9938
8911
74280
8028
9968
9004
69
307
9182
9999
9658
4(824)
93315
8001
9928
9196
102
338
8438
9967
9172
92368
9468
9999
9613
5(1030)
120
413
8041
9939
9449
133
443
8447
9976
9272
121
479
9544
9999
9752
6(1236)
155
476
8084
9877
9554
167
508
8863
9991
9569
159
536
9456
9999
9734
7(14
42)
195
573
7876
9892
9663
213
609
9124
9998
9729
205
643
9314
10000
9768
8(1648)
232
687
7964
9997
9788
267
724
8523
9988
9687
238
772
9422
9998
9689
9(1854)
293
692
8267
9893
9809
327
730
8386
9999
9750
298
757
9228
9999
9755
10(2060)
356
796
7946
9899
9801
399
840
8299
9976
9686
360
885
9668
9999
9705
11(2266)
423
829
8201
9895
9756
489
877
8378
9987
9788
448
927
9476
9999
9812
12(2472)
512
1023
8338
9913
9853
554
1085
8501
9999
9881
514
1148
9398
9999
9888
13(2678)
592
1045
8416
9907
9873
687
1088
7998
9996
9802
607
1152
9504
9999
9821
14(2884)
723
1099
7946
9915
9788
838
1166
8056
9992
9829
776
1230
9546
9999
9839
15(3090)
855
1089
7915
9964
9703
942
1156
8432
9984
9778
903
1225
9223
9998
9785
16(3296)
1009
1158
8024
9897
9769
1163
1231
8500
9996
9824
1045
1222
9199
9999
9835
17(34102)
1165
1299
8034
9919
9779
1285
1375
8495
9937
9808
1190
1457
9258
10000
9846
18(36108)
1254
1411
8176
9943
9885
1479
1492
8561
9968
9868
1345
1579
9286
9999
9875
19(38114
)1504
1506
7970
9923
9842
1651
1594
8167
9972
9846
1564
1694
9376
9999
9786
20(40120)
1706
1635
7819
9916
9783
1854
1734
8607
9968
9776
1778
1825
9285
9904
9718
21(42126)
2054
1622
8354
9918
9834
2357
1720
8297
9978
9833
2226
1856
9501
9999
9834
22(44132)
2400
1736
8494
9922
9846
2852
1844
8269
9999
9816
2680
1946
9486
9999
9820
23(46138)
2651
1934
8141
9953
9793
3159
2049
8492
9969
9799
3072
2172
9358
10000
9813
24(48144)
3004
1961
8226
9947
9681
3522
2086
8513
9998
9706
3102
2209
9488
9999
9724
25(50150)
3308
1924
8111
9965
9746
3898
2038
8264
9999
9788
3639
2161
9546
9999
9813
26(52156)
3669
1911
8096
9939
9855
4305
2027
8722
9967
9874
4453
2142
9185
9999
9876
27(54162)
4418
2095
8387
9927
9754
4771
2219
8341
9989
9749
4806
2353
9289
9999
9755
28(56168)
5367
2113
8458
9968
9848
5356
2237
8452
9978
9855
5221
2369
9228
10000
9868
29(58174)
6217
2186
7998
9989
9694
5942
2315
8316
9996
9699
5853
2371
9167
9999
9705
30(60180)
7263
2382
8321
9991
9837
6527
2543
8359
9998
9886
6214
2678
9369
9997
9886
31(62186)
8214
2346
8404
9936
9888
7813
2430
8321
9999
9846
6876
2560
9358
9999
9847
32(64192)
8863
2518
8378
9938
9807
8106
2677
8356
9999
9833
7877
2822
9408
10000
9835
33(66198)
9985
2532
8406
9917
9709
9684
2689
8497
9996
9746
8088
2838
9507
9999
9749
34(68204)
11264
2534
8543
9908
9897
10986
2685
8809
9996
9782
8469
2843
8844
9998
9780
35(70210)
12035
2715
8258
9937
9799
12098
2886
8569
9991
9809
9946
3041
9459
9999
9813
36(72216)
12803
2825
8104
9929
9827
13076
3010
8651
9994
9834
11013
3170
9468
9999
9865
37(74222)
13021
2830
8228
9949
9826
13261
3009
8376
9999
9826
12014
3172
9346
9999
9888
38(76228)
13862
2843
8152
9933
9807
13995
3028
8760
9998
9841
12133
3187
9706
9999
9876
39(78234)
15387
2953
8312
9964
9799
14827
3133
9001
9995
9783
13381
3304
8913
9999
9799
40(80240)
16974
2975
8510
9985
9767
15256
3249
8217
9993
9781
15867
3347
9111
9999
9791
Average
4352
15875
8143
9934
9708
4362
16856
8437
9985
9707
3968
17763
9356
9997
9781
10 Journal of Advanced Transportation
Table6Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mno
rmaldistr
ibution
set ( |119868 |
|119869 |)120572=
085120572=
09120572=
095Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
1(26)
27
907294
9964
9117
27
947844
9970
9202
2558
106
8422
10000
9208
2(412
)46
170
7961
9921
9283
48
183
8328
9982
9283
4508
183
8823
10000
9370
3(618
)68
258
7578
9916
9228
68
280
7798
9971
9417
6803
307
7658
9998
9447
4(824)
93315
8000
9899
9367
96337
8379
9979
9467
9373
368
8438
9999
9579
5(1030)
120
413
7841
9927
9437
124
443
8763
9963
9476
1262
499
8850
9998
9708
6(1236)
155
476
7921
9919
9519
162
504
8124
9982
9582
16021
536
8954
10000
9756
7(14
42)
195
573
7976
9933
9598
196
609
8287
9975
9619
2011
646
8467
9999
9767
8(1648)
232
687
8031
9918
9705
235
727
8115
9965
9650
2404
772
8278
9999
9765
9(1854)
293
692
8086
9898
9759
291
734
8464
9974
9755
29815
796
8617
9998
9835
10(2060)
356
796
7984
9937
9770
358
842
8223
9988
9765
37973
921
8846
10000
9838
11(2266
)421
829
8101
9940
9755
436
877
8211
9986
9786
4514
8959
8497
9999
9789
12(2472)
514
1023
8033
9951
9785
501
1085
8162
9983
9788
52073
1147
8321
9999
9824
13(2678)
592
1023
8138
9916
9885
599
1087
8411
9970
9845
62937
1194
8609
9998
9795
14(2884)
728
1099
8235
9913
9915
722
1168
8435
9987
9806
75053
1270
8567
9997
9834
15(3090)
855
1089
8187
9922
9861
880
1157
8376
9976
9796
89222
1300
8611
9996
9798
16(3296)
1009
1158
8770
9897
9788
966
1231
8845
9984
9823
106288
1359
8997
9999
9848
17(34102)
1154
1299
8434
9925
9849
1137
1377
8601
9980
9798
123869
1493
8798
9996
9877
18(36108)
1407
1411
8321
9927
9866
1339
1495
8301
9965
9822
133807
1601
8567
9999
9834
19(38114
)1702
1506
8206
9915
9805
1672
1598
8977
9978
9845
152999
1694
8998
9999
9847
20(40120)
2091
1635
8528
9918
9889
1916
1737
8805
9976
9837
1890
021833
9012
9999
9791
21(42126)
2843
1622
8550
9934
9913
2432
1718
8886
9974
9813
2399
131815
9008
10000
9815
22(44132)
3300
1736
8449
9940
9859
2795
1843
8786
9977
9756
253106
1948
8997
9999
9826
23(46138)
3741
1934
8133
9911
9876
3143
2050
8876
9981
9785
274583
2173
9014
9999
9858
24(48144)
4531
1961
8245
9909
9846
3515
2090
8344
9993
9819
308089
2211
8661
9999
9883
25(50150)
4715
1924
8111
9929
9888
3959
2041
8279
9996
9837
3510
932158
8527
9999
9827
26(52156)
5223
1911
8096
9898
9857
4301
2025
8251
9997
9757
421747
2146
8497
9999
9914
27(54162)
5994
2095
8387
9908
9775
4835
2226
8497
9974
9780
4614
432359
8695
9999
9837
28(56168)
6406
2111
8411
9906
9798
5336
2240
8558
9991
9820
494113
2371
9249
9999
9816
29(58174)
6897
2176
8012
9916
9791
5948
2311
8211
9984
9765
624835
2445
8531
10000
9851
30(60180)
7436
2391
8111
9913
9766
6558
2533
8305
9968
9795
696116
2682
8575
9999
9799
31(62186)
8118
2282
8119
9928
9806
7069
2422
8417
9973
9873
797768
2564
8667
9999
9846
32(64192)
8895
2519
8223
9919
9757
7857
2675
8631
9982
9769
864813
2827
8976
10000
9905
33(66198)
11235
2533
8563
9928
9811
9162
2688
8547
9979
9828
973089
2843
8657
9999
9846
34(68204)
11587
2534
8249
9926
9847
9549
2686
8798
9938
9856
10523
3041
8999
10000
9877
35(70210)
1164
92715
8098
9895
9872
10345
2880
8602
9998
9869
11272
3052
9021
9999
9897
36(72216)
12224
2825
8403
9958
9770
11165
3002
8657
9999
9839
12456
3175
9113
9999
9864
37(74222)
12978
2830
8357
9937
9795
11850
3005
8496
9998
9894
13458
3176
8827
9999
9905
38(76228)
13426
2844
8267
9966
9759
12795
3019
8604
9999
9797
15155
3121
8978
9999
9853
39(78234)
15003
2955
8303
9956
9867
13816
3132
8588
9996
9885
15041
3310
9016
9999
9912
40(80240)
16495
2978
8257
9954
9826
14896
3160
8495
9997
9837
16563
3349
8957
9999
9933
Average
4619
15855
8174
9925
9742
4077
16828
8457
9981
9743
4350
17938
8732
9999
9794
Journal of Advanced Transportation 11
high service level requirement we recommend MIP-DF1 forsolutions
6 Conclusion
This paper investigates an ambulance location problem withambiguity set of demand in which the probability distribu-tion of the uncertain demand is unknown In such a data-driven environment only the mean and covariance demandsat emergency points are knownThe problem is to determinethe base locations and the number of ambulances at eachbase aiming at minimizing the total cost associated withlocating bases and assigning ambulances We propose adistribution-free model with chance constraints Then twoapproximated MIP formulations with different approxima-tion methods of chance constraints are proposed which arebased on different ambiguity sets of the unknown probabilitydistribution Numerical experiments on benchmarks and120 randomly generated instances are conducted and thecomputational results show that the first approximated MIPformulation is much conservative with overhigh system costwhile the second approximatedMIP formulation ismore costsaving with service level appropriately ensured
In future research wemay consider the following relevantissues First the uncertainty of driving time from a base toan emergency point shall be analysed Besides unexpected orsudden disasters may be taken into consideration Moreoverother uncertainties should be considered such as travel timethe number of available vehicles and so on The stochasticdependence should also further considered
Data Availability
The data of the benchmark instances in our manuscript canbe obtained in Nickel et al [8] For the test instances thedata is randomly generated and the way for generating data isstated in our manuscript The datasets generated during thecurrent study are available from the corresponding author onreasonable request
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This work was supported by the National Natural Sci-ence Foundation of China (NSFC) under Grants 7142800271531011 71571134 and 71771048 This work was also sup-ported by the Fundamental Research Funds for the CentralUniversities
References
[1] L Brotcorne G Laporte and F Semet ldquoAmbulance loca-tion and relocation modelsrdquo European Journal of OperationalResearch vol 147 no 3 pp 451ndash463 2003
[2] X Li Z Zhao X Zhu and T Wyatt ldquoCovering modelsand optimization techniques for emergency response facilitylocation and planning a reviewrdquo Mathematical Methods ofOperations Research vol 74 no 3 pp 281ndash310 2011
[3] C Toregas R Swain C ReVelle and L Bergman ldquoThe locationof emergency service facilitiesrdquoOperations Research vol 19 no6 pp 1363ndash1373 1971
[4] R Church and C ReVelle ldquoThe maximal covering locationproblemrdquo Papers in Regional Science vol 32 no 1 pp 101ndash1181974
[5] M Gendreau G Laporte and F Semet ldquoSolving an ambulancelocation model by tabu searchrdquo Location Science vol 5 no 2pp 75ndash88 1997
[6] A Ingolfsson S Budge and E Erkut ldquoOptimal ambulancelocation with random delays and travel timesrdquo Health CareManagement Science vol 11 no 3 pp 262ndash274 2008
[7] M Gendreau G Laporte and F Semet ldquoThemaximal expectedcoverage relocation problem for emergency vehiclesrdquo Journal ofthe Operational Research Society vol 57 no 1 pp 22ndash28 2006
[8] S Nickel M Reuter-Oppermann and F Saldanha-da-GamaldquoAmbulance location under stochastic demand A samplingapproachrdquo Operations Research for Health Care vol 8 pp 24ndash32 2016
[9] P Beraldi M E Bruni and D Conforti ldquoDesigning robustemergency medical service via stochastic programmingrdquo Euro-pean Journal of Operational Research vol 158 no 1 pp 183ndash1932004
[10] N Noyan ldquoAlternate risk measures for emergency medicalservice system designrdquo Annals of Operations Research vol 181pp 559ndash589 2010
[11] M R Wagner ldquoStochastic 0-1 linear programming underlimited distributional informationrdquoOperations Research Lettersvol 36 no 2 pp 150ndash156 2008
[12] EDelage andYYe ldquoDistributionally robust optimization undermoment uncertainty with application to data-driven problemsrdquoOperations Research vol 58 no 3 pp 595ndash612 2010
[13] J T Essen J L Hurink S Nickel and M Reuter ldquoModels forambulance planning on the strategic and the tactical levelrdquo BetaResearch School for Operations Management and Logistics p 272013
[14] E Erkut A Ingolfsson and G Erdogan ldquoAmbulance locationfor maximum survivalrdquoNaval Research Logistics (NRL) vol 55no 1 pp 42ndash58 2008
[15] S C Chapman and J A White ldquoProbabilistic formulations ofemergency service facilities location problemsrdquo in Proceedingsof ORSATIMS Conference San Juan Puerto Rico 1974
[16] A Ben-Tal D Den Hertog A De Waegenaere B Melenbergand G Rennen ldquoRobust solutions of optimization problemsaffected by uncertain probabilitiesrdquo Management Science vol59 no 2 pp 341ndash357 2013
[17] MNg ldquoDistribution-free vessel deployment for liner shippingrdquoEuropean Journal of Operational Research vol 238 no 3 pp858ndash862 2014
[18] M Ng ldquoContainer vessel fleet deployment for liner shippingwith stochastic dependencies in shipping demandrdquo Transporta-tion Research Part B Methodological vol 74 pp 79ndash87 2015
[19] R Jiang and Y Guan ldquoData-driven chance constrained stochas-tic programrdquo Mathematical Programming vol 158 no 1-2 SerA pp 291ndash327 2016
[20] C-M Lee and S-L Hsu ldquoThe effect of advertising on thedistribution-free newsboy problemrdquo International Journal ofProduction Economics vol 129 no 1 pp 217ndash224 2011
[21] K Kwon and T Cheong ldquoA minimax distribution-free proce-dure for a newsvendor problem with free shippingrdquo EuropeanJournal of Operational Research vol 232 no 1 pp 234ndash2402014
12 Journal of Advanced Transportation
[22] Y Zhang S Shen and S A Erdogan ldquoDistributionally robustappointment scheduling with moment-based ambiguity setrdquoOperations Research Letters vol 45 no 2 pp 139ndash144 2017
[23] Y Zhang S Shen and S A Erdogan ldquoSolving 0-1semidefinite programs for distributionally robust allocationof surgery blocksrdquo Social Science Electronic Publishing 2017httpsssrncomabstract=2997524
[24] F Zheng J He F Chu and M Liu ldquoA new distribution-freemodel for disassembly line balancing problem with stochastictask processing timesrdquo International Journal of ProductionResearch pp 1ndash13 2018
[25] J Cheng E Delage and A Lisser ldquoDistributionally robuststochastic knapsack problemrdquo SIAM Journal on Optimizationvol 24 no 3 pp 1485ndash1506 2014
[26] M Liu D Yang and F Hao ldquoOptimization for the locationsof ambulances under two-stage life rescue in the emergencymedical service a case study in Shanghai ChinardquoMathematicalProblems in Engineering vol 2017 Article ID 1830480 14 pages2017
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
Journal of Advanced Transportation 5
Table 1 120572 = 095|119868| = |119869| Sampling approach MIP-DF1 MIP-DF2
CT Obj Sel () CT Obj Sel () CT Obj Sel ()1 7 49 23 9626 01 36 10000 01 23 100002 7 51 22 9863 01 35 10000 01 24 99823 6 41 14 9692 01 30 10000 01 14 98504 6 43 13 9713 02 29 10000 01 12 97425 6 57 12 7996 01 30 10000 02 12 100006 8 58 21 9850 01 24 10000 02 27 10000
Average 50 175 9457 01 307 10000 01 187 9929
The second approximated MIP formulationMIP-DF2 toDF is presented as follows
[MIP-DF2]
min sum119894isin119868
(119891119894 sdot 119909119894 + 119892119894 sdot 119911119894)st Constraints (4) - (8) (15)
(16)
Notice that when 1205741 = 0 and 1205742 = 1 inequality (14)reduces to (11) in MIP-DF1 which implies that MIP-DF1 isa special case of MIP-DF2
5 Computational Experiments
In this section the performance of our proposed formula-tions first evaluated and compared to the sampling approachproposed in Nickel et al [8] Specifically given a requiredservice level 120572 we apply our MIP formulations by restrictingthe chance constraint for each emergency point to satisfya safety level 120572119895 = 120572 while we adopt the samplingapproach by restricting the coverage constraint with 120572 Thencomputational results on 120 randomly generated instancesbased on the information on emergencies in Shanghai arereported Our approximated MIP formulations and thesampling approach are solved by coding in MATLAB 2014bcalling CPLEX 126 solver All numerical experiments areconducted on a personal computer with Core I5 and 330GHzprocessor and 8GB RAMunder windows 7 operating systemThe computational time for sampling approach is limited to3600 seconds
51 Out-of-Sample Test FollowingZhang et alrsquos [22]method-ology we examine the out-of-sample performance of solu-tions obtained by our proposed approximated MIP formula-tions and the sampling approach [8] The out-of-sample testincludes three steps
(1) The base locations and the employment of ambu-lances at each base are determined with known meanvector and covariance matrix of demands which isnamed as in-sample test
(2) 1000 scenarios are sampled from the underlying truedistribution which represent realizing the uncertaindemand
(3) Based on the solution obtained in step (1) ambu-lances are assigned to serve emergency points inthe 1000 scenarios to maximize the (out-of-sample)service level
The (out-of-sample) service level is defined as the ratioof the number of emergency points with demand satisfiedto the total number of emergency points in 1000 scenariosIn the following subsections the benchmarks are tested andthen three common distributions are employed ie uniformdistribution Poisson distribution and normal distributionrespectively
52 Benchmark Instances We first test the benchmarks inNickel et al [8] to compare our proposed two approximatedMIP formulationswith the sampling approach InNickel et al[8] they assume that demand nodes are also potential baselocations and the number of ambulances required by eachemergency point is chosen from set 0 1 2with given proba-bility distribution Computational results on benchmarks arereported in Tables 1 and 2
Table 1 reports the computational results with servicelevel requirement120572 = 095 In Table 1 CTObj and Sel denotethe computational time objective value (ie the system cost)and the service level respectively The first column includesthe indexes of instances and the second indicates the numberof nodes It shows that the average service level obtainedby the sampling approach is 9457 which is smaller thanthe required service level ie 85 Specifically the servicelevel of instance 5 obtained by the sampling approach is only7996 However MIP-DF1 and MIP-DF2 can guarantee theservice levels for all instances higher than the requirementConcluding from Table 1 we can observe that (i) the averagecomputational time of sampling approach is about 50 timesas those of the two approximated MIP formulations and (ii)MIP-DF1 and MIP-DF2 improve the average service level byabout 574 and 499 compared with sampling approachhowever (iii) system costs obtained by MIP-DF1 and MIP-DF2 are respectively 7543 and 686 higher than that ofsampling approach
Table 2 reports the computational results with servicelevel requirement 120572 = 099 For instances 2-6 the ser-vice levels obtained by sampling approach cannot meet therequirement ie 99 Besides it shows that the averagecomputational time of sampling approach is 48 which isabout 48 times greater than those of MIP-DF1 andMIP-DF2
6 Journal of Advanced Transportation
Table 2 120572 = 099|119868| = |119869| Sampling approach MIP-DF1 MIP-DF2
CT Obj Sel () CT Obj Sel () CT Obj Sel ()1 7 49 24 9979 01 64 10000 01 25 100002 7 48 23 9856 01 63 10000 01 24 100003 6 51 15 9717 01 55 10000 01 16 100004 6 47 15 9275 01 52 10000 01 13 100005 6 40 14 9000 01 55 10000 01 14 100006 8 52 23 9785 01 45 10000 01 27 10000
Average 48 190 9343 01 557 10000 01 198 10000
Table 3 Solutions to instance 4 in Nickel et al [8] for the given scenario
Sampling approach MIP-DF1 MIP-DF2Emergency points 1 2 3 4 5 6 Emergency points 1 2 3 4 5 6 Emergency points 1 2 3 4 5 6
Demand 1 0 2 0 2 0 Demand 1 0 2 0 2 0 Demand 1 0 2 0 2 0Base locations Ambulance employment Ambulance employment Ambulance employment1 0 0 02 1 1 0 3 1 23 1 1 26 1 2 2 04 0 0 05 0 6 06 2 2 0 4 2
The average service level is 9343 Although the averageservice levels obtained by MIP-DF1 and MIP-DF2 are both100 the average system cost obtained by MIP-DF1 is about3 times larger than that of sampling approach Besides theaverage system cost obtained by MIP-DF2 is 421 higherthan that of sampling approach
Concluding for the benchmarks described in Nickel etal [8] we can observe that (1) MIP-DF1 and MIP-DF2 canbe solved in far less computational time compared with thesampling approach (2) the sampling approach cannot meetthe requirement of service level for all benchmarks whilethe two MIP formulations can (3) the average system cost ofMIP-DF2 is very close to that of sampling approach (4) withthe increase of 120572 the average system costs obtained by MIP-DF2 is getting closer to that obtained by sampling approach
From Tables 1 and 2 we can observe that in spite ofthe lower system cost obtained by the sampling approachthere may exist extreme situations that demands of someemergency points are not satisfied Take one benchmark inNickel et al [8] ie instance 4 where there are 6 nodesA scenario in which demand at each emergency point isgenerated from the given probability distribution is shown inFigure 1 We can observe from Figure 1 that under the givenscenario the demand of each possible emergency point is1 0 2 0 2 0The solutions obtained by sampling approachMIP-DF1 andMIP-DF2 are shown in Table 3 which includesbase locations and ambulance employment obtained by in-sample test and ambulance assignment under given scenario
In Table 3 the first column includes the indexes of poten-tial base locations The second column represents the baselocations and the ambulance employment at each location
>1 = 1
>2 = 0
>3 = 2
>4 = 0
>5 = 2
>6 = 0
1
2 4
3 5
6
Figure 1 Instance 4 in Nickel et al [8] under given scenario
in which a location with 0 ambulance implies that it is notselected for base construction From Table 3 we can observethat the in-sample solutions obtained by sampling approachis to construct bases at locations 2 3 and 6 where the numberof ambulances is 1 1 and 2 respectively Besides undergiven scenario there is one ambulance serving emergencypoint 1 from base location 3 and one ambulance servingemergency point from base location 2 and 2 ambulancesserving emergency point 5 from base location 6 It shows thatthe requirement of ambulances at emergency point 3 is notfully satisfied The insufficient service may cause serious lossof life However we can find that the solutions obtained byMIP-DF1 andMIP-DF2 can provide sufficient ambulances forall emergency points
In sum to guarantee the required service level withless cost MIP-DF2 is recommended for the solutions with
Journal of Advanced Transportation 7
Nickel et al [8]rsquos data set Moreover in the benchmarks itis assumed that all nodes are possible emergency points aswell as potential base locations and the number of requiredambulances is small which is not always consistent withthe situation in Shanghai Therefore we further test the twoapproximated MIP formulations and the sampling approachon randomly generated instances by analysing the informa-tion on emergencies in Shanghai [26]
53 Computational Tests on Randomly Generated InstancesIn the following subsection numerical experiments on 120randomly generated instances are described
531 Test Instances By analysing the information on emer-gencies in Shanghai [26] the mean value 120583119895 of demand 119889119895 ineach emergency point are randomly generated fromadiscreteUniform distribution on interval [5 25] The standard devia-tion 120590119895 is set to be 5 Since the sampling approach proposed inNickel et al [8] requires a sample of scenarios in numericalexperiments scenarios are produced with demands randomlygenerated from Uniform distribution Poisson distributionand Normal distribution consistent with given mean andstandard deviation
According to the instance settings in Erkut and Ingolfsson(2008) response time from a base to an emergency pointis randomly generated from a discrete Uniform distributionon interval [3 30] in units of minutes A base covers anemergency point if the driving time between them is lessthan 10 minutes to ensure the survival probability (Erkut andIngolfsson 2006) For an emergency point in each instanceif there is no base with 10 minutes driving time to it then thebase covering this point is set to be the nearest one The costof constructing a base and that of locating one ambulance areboth set to be 1 as in Nickel et al [8] Besides the number ofemergency points is 3 times that of potential base locations
532 Results and Discussion Different values of requiredservice level 120572 are tested ie 085 09 095 Computationalresults on 120 randomly generated instances are reported inTables 4-6
Table 4 reports the numerical results of instances withdemand generated from Uniform distribution We observethat when 120572 = 085 09 095 the average computationaltimes of the sampling approach are 4482 4379 and 4295seconds respectively However both approximated MIP for-mulations can produce solutions in much faster speeds iewithin 13 seconds on average With the increase of 120572 (i)the out-of-sample average service level is getting higher and(ii) the system costs of solutions obtained by all method aregetting larger and (iii) the system costs of solutions yieldedbyMIP-DF2 is getting closer to those of solutions obtained bythe sampling approach In terms of the average system costthe sampling approach performs the best while MIP-DF1 isthe poorest When 120572 = 085 for example the average systemcost ofMIP-DF2 and the sampling approach are about 3032and 442 less than that of MIP-DF1 respectively When120572 = 095 the above gaps are even larger Concluding fortheUniformdistribution (1) the proposed two approximatedMIP formulations are far more time saving than the sampling
approach (2)MIP-DF1 andMIP-DF2 can improve the servicelevel by about 1907 compared with the sampling approachhowever (3) despite of the highest service levels obtainedby MIP-DF1 the system cost obtained by MIP-DF1 is alsohigh (4) MIP-DF2 can ensure an appropriate service levelwith much less cost than MIP-DF1 and (5) for the situationswith small-scale instances and high service level requirementMIP-DF1 is recommended
Computational results of instances with demand at eachemergency point generated from Poisson distribution arepresented in Table 5 For MIP-DF1 and MIP-DF2 onlymean and variance are involved Therefore with givenmean and standard deviation of the demand at each pointcomputational results are the same for different probabilitydistributions We can obtain from Table 5 that the objectivevalue obtained by the sampling approach increases with theincrease of 120572 Besides when the value of 120572 equals 085the average service level obtained by the sampling approachis about 8143 which is smaller than that of the twoapproximated MIP formulations The highest service level9934 is obtained by MIP-DF1 and that obtained by MIP-DF2 is 9708 When 120572 = 09 the average service levelof the sampling approach is 8437 and those obtained byMIP-DF1 and MIP-DF2 are 9985 and 9797 respectivelyWhen 120572 = 095 the average service level obtained by thesampling approach is 9356 and those yielded by MIP-DF1 and MIP-DF2 are 9997 and 9781 respectively ForPoisson distribution concluding MIP-DF1 and MIP-DF2can improve the service level compared with the samplingapproach especially MIP-DF1 can obtain the highest servicelevel
Similarly Table 6 reports the computational results forinstances in which the demand at each emergency point isgenerated following Normal distribution When 120572 = 085the average service levels obtained by the sampling approachMIP-DF1 and MIP-DF2 are 8174 9925 and 9742respectively When 120572 = 09 the three values are 84579981 and 9743 respectively Moreover for 120572 = 095 thevalues are equal to 8732 9999 and 9794 respectivelyIt can be observed that the performance of the samplingapproach is the poorest in terms of the service level
By observing the above computational results on bench-marks and randomly generated instances we conclude that(1)The proposed approximated MIP formulations ie MIP-DF1 and MIP-DF2 are very time saving compared with thesampling approach (2) MIP-DF1 an MIP-DF2 can achievethe requirement of service level and improve the service levelby about 1547 and 1229 on average compared with thatof the sampling approach (3) MIP-DF1 is more conservativein terms of overhigh emergency service level close to 100and the system cost is also high which is about 568 largerthan that of MIP-DF2 (4) Compared with MIP-DF1 MIP-DF2 is less conservative and much more cost saving whileensuring an appropriately service level on average which isabout 97
In sum for medium and large problem instances (egShanghai) which are common in practice to guarantee therequired service level with less cost we recommendMIP-DF2as solution method However for very small instances with
8 Journal of Advanced Transportation
Table4Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mun
iform
distrib
ution
Set (|119868||119869|)
120572=085
120572=09
120572=095
Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
1(26)
27
897317
01
149
9906
01
107
9222
28
967373
01
155
9933
01
107
9167
27
106
7585
01
185
9987
01
107
9128
2(412
)47
169
7423
00
291
9919
00
207
9122
58
181
7438
01
303
9997
01
207
9363
50
199
7702
00
363
9982
00
207
9338
3(618
)72
258
7506
01
438
9976
01
312
9106
75279
8082
01
456
9996
01
312
9369
78304
8337
01
546
9997
01
312
9357
4(824)
99314
7956
01
553
9993
01
385
9206
111
338
8202
01
577
9999
01
385
9602
112
366
8492
01
697
9994
01
385
9620
5(1030)
132
415
8406
01
722
9896
01
512
9406
143
441
8322
02
752
9983
01
512
9692
142
476
8609
01
902
9993
01
512
9692
6(1236)
170
475
8249
02
843
9870
01
591
9367
181
504
8251
01
879
9982
01
591
9608
189
535
8726
01
1059
9997
01
591
9609
7(14
42)
215
573
8111
02
1006
9849
02
712
9606
243
609
8217
02
1048
9953
02
712
9589
233
644
8608
02
1258
9992
02
712
9582
8(1648)
248
686
7532
02
1185
9896
02
849
9725
284
727
8307
02
1233
9949
02
849
9613
281
768
9063
02
1473
9999
03
849
9622
9(1854)
313
691
8353
03
1236
9974
04
858
9709
351
731
8397
03
1290
9988
03
858
9765
352
760
8906
03
1560
9994
03
858
9758
10(2060)
372
794
8174
03
1409
9887
03
989
9657
428
843
8489
03
1469
9947
03
989
9773
432
892
9004
03
1769
9997
03
989
9738
11(2266
)451
830
8072
04
1493
9949
03
1031
9758
518
880
8328
04
1559
9982
03
1031
9723
516
930
8967
05
1889
9995
03
1031
9727
12(2472)
543
1025
7970
04
1771
9927
04
1267
9835
601
1083
8482
05
1843
9949
04
1267
9776
637
1146
9014
04
2203
9993
04
1267
9777
13(2678)
631
1026
8142
05
1819
9898
04
1274
9782
743
1087
8678
05
1897
9936
04
1274
9778
765
1146
9023
04
2287
9997
05
1273
9837
14(2884)
782
1101
7903
05
1949
9869
08
1361
9804
878
1165
8790
05
2033
9935
07
1361
9889
870
1227
9056
05
2453
9998
06
1361
9808
15(3090)
907
1091
8098
06
1997
9915
06
1367
9629
1018
1155
8329
06
2087
9981
06
1367
9807
1020
1225
8986
06
2537
9999
08
1367
9817
16(3296)
1062
1160
8798
07
2123
9944
07
1449
9787
1183
1229
8383
07
2219
9988
06
1449
9849
1193
1300
9105
06
2698
9998
06
1449
9815
17(34102)
1202
1298
7572
07
2328
9969
07
1614
9821
1377
1374
8860
08
2430
9997
07
1614
9824
1260
1452
9085
07
2940
9999
08
1614
9818
18(36108)
1394
1408
8121
08
2509
9947
08
1752
9809
1544
1492
8928
08
2617
9983
08
1752
9799
1548
1580
9173
08
3156
9999
08
1752
9809
19(38114
)1618
1508
8170
09
2676
9899
111877
9742
1760
1598
9130
09
2790
9934
08
1877
9908
1755
1688
9186
09
3360
9999
09
1877
9872
20(40120)
2206
1628
7719
102869
9989
102029
9873
2007
1732
8753
112990
9997
09
2029
9879
1904
1827
9138
103590
9998
09
2029
9875
21(42126)
2780
1622
8354
102900
9978
102017
9732
2609
1718
8854
103026
9876
102017
9818
2670
1812
9089
123655
9999
102017
9848
22(44132)
3189
1735
8494
113088
9985
152165
9857
2970
1840
8922
123220
9925
122165
9886
2873
1952
9206
113880
9999
112165
9850
23(46138)
3630
1937
8141
123372
9893
132406
9796
3222
2048
8937
123510
9934
142406
9906
3461
2170
8989
124199
10000
122406
9809
24(48144)
4226
1968
8226
133461
9932
132453
9814
3674
2085
8846
133605
9956
152453
9912
3556
2211
9300
134326
9999
132453
9819
25(50150)
4727
1919
8111
143456
9999
152407
9629
4019
2037
8699
143606
9999
142407
9857
4257
2161
9259
154356
9995
142407
9819
26(52156)
5300
1906
8096
153484
9878
172392
9658
4495
2021
8952
143640
9999
152392
9568
4486
2142
9182
1544
209999
152392
9854
27(54162)
5913
2096
8387
163751
9846
172617
9761
5117
2221
9045
163913
9998
172617
9544
4830
2355
9346
184723
9999
162617
9893
28(56168)
6310
2110
7758
173819
9933
192643
9829
5777
2234
9157
183987
9998
182643
9526
5550
2365
9045
184827
9999
172643
9831
29(58174)
6780
2175
7998
183937
9898
182720
9714
6303
2307
9022
20
4111
9969
192720
9623
6189
2444
9128
194981
10000
192720
9914
30(60180)
7593
2384
8321
194240
9967
20
2979
9841
6791
2527
8978
1944
199979
23
2979
9721
6878
2679
9143
21
5316
9999
20
2979
9869
31(62186)
8220
2281
7604
20
4157
9984
21
2856
9796
7734
2416
8934
21
4343
9997
21
2856
9679
7466
2557
9158
22
5273
9999
23
2856
9902
32(64192)
8766
2517
8378
21
4489
9932
25
3145
9729
8276
2671
8296
23
4680
9977
24
3145
9825
7897
2822
9089
24
5640
9999
22
3145
9891
33(66198)
10123
2532
7806
23
4555
9967
23
3169
9615
9010
2682
8947
24
4752
9989
24
3169
9703
9051
2835
9157
24
5742
9998
24
3169
9867
34(68204)
11010
2535
7934
24
4602
9999
29
3174
9631
9898
2685
8858
30
4806
9998
28
3174
9762
9863
2844
9236
26
5824
9999
25
3174
9889
35(70210)
11666
2714
8062
25
4862
9986
29
3391
9787
1066
62875
9137
26
5072
9999
29
3391
9716
10822
3043
9315
26
6122
9999
29
3391
9923
36(72216)
12011
2827
8104
26
5047
9991
28
3535
9798
1246
52984
9016
28
5263
9999
28
3535
9878
11667
3169
9253
30
6343
9999
29
3535
9896
37(74222)
11437
2826
8228
28
5096
9933
29
3543
9884
13582
3098
8895
29
5318
9999
32
3543
9888
12654
3178
9139
29
6427
10000
31
3543
9893
38(76228)
12435
2839
8152
29
5161
9897
29
3565
9788
13976
3103
8926
29
5388
9999
31
3565
9724
13837
3185
9255
33
6528
10000
31
3565
9895
39(78234)
1464
32945
8412
31
5331
9998
32
3695
9876
14794
3210
9005
31
5565
9999
34
3695
9713
14561
3313
9081
34
6735
10000
33
3695
9878
40(80240)
1604
22975
8510
32
5414
9957
36
3734
9796
16263
3346
8999
35
5654
9999
35
3734
9806
15869
3343
9248
35
6854
10000
35
3734
9873
Average
4482
1584
680
67
1228
397
9936
131978
796
82
4379
16913
8654
1329
626
9975
131978
79721
4295
1778
889
85
1335774
9997
131978
79775
Journal of Advanced Transportation 9
Table5Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mPo
isson
distr
ibution
set ( |119868 |
|119869 |)120572=
085120572=
09120572=
095Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
1(26)
27
907294
9967
9001
27
987770
9979
9013
24
103
9273
9999
9468
2(412
)46
170
7961
9921
9117
51
183
8305
9946
9128
49
201
9392
9998
9546
3(618
)68
258
7578
9938
8911
74280
8028
9968
9004
69
307
9182
9999
9658
4(824)
93315
8001
9928
9196
102
338
8438
9967
9172
92368
9468
9999
9613
5(1030)
120
413
8041
9939
9449
133
443
8447
9976
9272
121
479
9544
9999
9752
6(1236)
155
476
8084
9877
9554
167
508
8863
9991
9569
159
536
9456
9999
9734
7(14
42)
195
573
7876
9892
9663
213
609
9124
9998
9729
205
643
9314
10000
9768
8(1648)
232
687
7964
9997
9788
267
724
8523
9988
9687
238
772
9422
9998
9689
9(1854)
293
692
8267
9893
9809
327
730
8386
9999
9750
298
757
9228
9999
9755
10(2060)
356
796
7946
9899
9801
399
840
8299
9976
9686
360
885
9668
9999
9705
11(2266)
423
829
8201
9895
9756
489
877
8378
9987
9788
448
927
9476
9999
9812
12(2472)
512
1023
8338
9913
9853
554
1085
8501
9999
9881
514
1148
9398
9999
9888
13(2678)
592
1045
8416
9907
9873
687
1088
7998
9996
9802
607
1152
9504
9999
9821
14(2884)
723
1099
7946
9915
9788
838
1166
8056
9992
9829
776
1230
9546
9999
9839
15(3090)
855
1089
7915
9964
9703
942
1156
8432
9984
9778
903
1225
9223
9998
9785
16(3296)
1009
1158
8024
9897
9769
1163
1231
8500
9996
9824
1045
1222
9199
9999
9835
17(34102)
1165
1299
8034
9919
9779
1285
1375
8495
9937
9808
1190
1457
9258
10000
9846
18(36108)
1254
1411
8176
9943
9885
1479
1492
8561
9968
9868
1345
1579
9286
9999
9875
19(38114
)1504
1506
7970
9923
9842
1651
1594
8167
9972
9846
1564
1694
9376
9999
9786
20(40120)
1706
1635
7819
9916
9783
1854
1734
8607
9968
9776
1778
1825
9285
9904
9718
21(42126)
2054
1622
8354
9918
9834
2357
1720
8297
9978
9833
2226
1856
9501
9999
9834
22(44132)
2400
1736
8494
9922
9846
2852
1844
8269
9999
9816
2680
1946
9486
9999
9820
23(46138)
2651
1934
8141
9953
9793
3159
2049
8492
9969
9799
3072
2172
9358
10000
9813
24(48144)
3004
1961
8226
9947
9681
3522
2086
8513
9998
9706
3102
2209
9488
9999
9724
25(50150)
3308
1924
8111
9965
9746
3898
2038
8264
9999
9788
3639
2161
9546
9999
9813
26(52156)
3669
1911
8096
9939
9855
4305
2027
8722
9967
9874
4453
2142
9185
9999
9876
27(54162)
4418
2095
8387
9927
9754
4771
2219
8341
9989
9749
4806
2353
9289
9999
9755
28(56168)
5367
2113
8458
9968
9848
5356
2237
8452
9978
9855
5221
2369
9228
10000
9868
29(58174)
6217
2186
7998
9989
9694
5942
2315
8316
9996
9699
5853
2371
9167
9999
9705
30(60180)
7263
2382
8321
9991
9837
6527
2543
8359
9998
9886
6214
2678
9369
9997
9886
31(62186)
8214
2346
8404
9936
9888
7813
2430
8321
9999
9846
6876
2560
9358
9999
9847
32(64192)
8863
2518
8378
9938
9807
8106
2677
8356
9999
9833
7877
2822
9408
10000
9835
33(66198)
9985
2532
8406
9917
9709
9684
2689
8497
9996
9746
8088
2838
9507
9999
9749
34(68204)
11264
2534
8543
9908
9897
10986
2685
8809
9996
9782
8469
2843
8844
9998
9780
35(70210)
12035
2715
8258
9937
9799
12098
2886
8569
9991
9809
9946
3041
9459
9999
9813
36(72216)
12803
2825
8104
9929
9827
13076
3010
8651
9994
9834
11013
3170
9468
9999
9865
37(74222)
13021
2830
8228
9949
9826
13261
3009
8376
9999
9826
12014
3172
9346
9999
9888
38(76228)
13862
2843
8152
9933
9807
13995
3028
8760
9998
9841
12133
3187
9706
9999
9876
39(78234)
15387
2953
8312
9964
9799
14827
3133
9001
9995
9783
13381
3304
8913
9999
9799
40(80240)
16974
2975
8510
9985
9767
15256
3249
8217
9993
9781
15867
3347
9111
9999
9791
Average
4352
15875
8143
9934
9708
4362
16856
8437
9985
9707
3968
17763
9356
9997
9781
10 Journal of Advanced Transportation
Table6Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mno
rmaldistr
ibution
set ( |119868 |
|119869 |)120572=
085120572=
09120572=
095Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
1(26)
27
907294
9964
9117
27
947844
9970
9202
2558
106
8422
10000
9208
2(412
)46
170
7961
9921
9283
48
183
8328
9982
9283
4508
183
8823
10000
9370
3(618
)68
258
7578
9916
9228
68
280
7798
9971
9417
6803
307
7658
9998
9447
4(824)
93315
8000
9899
9367
96337
8379
9979
9467
9373
368
8438
9999
9579
5(1030)
120
413
7841
9927
9437
124
443
8763
9963
9476
1262
499
8850
9998
9708
6(1236)
155
476
7921
9919
9519
162
504
8124
9982
9582
16021
536
8954
10000
9756
7(14
42)
195
573
7976
9933
9598
196
609
8287
9975
9619
2011
646
8467
9999
9767
8(1648)
232
687
8031
9918
9705
235
727
8115
9965
9650
2404
772
8278
9999
9765
9(1854)
293
692
8086
9898
9759
291
734
8464
9974
9755
29815
796
8617
9998
9835
10(2060)
356
796
7984
9937
9770
358
842
8223
9988
9765
37973
921
8846
10000
9838
11(2266
)421
829
8101
9940
9755
436
877
8211
9986
9786
4514
8959
8497
9999
9789
12(2472)
514
1023
8033
9951
9785
501
1085
8162
9983
9788
52073
1147
8321
9999
9824
13(2678)
592
1023
8138
9916
9885
599
1087
8411
9970
9845
62937
1194
8609
9998
9795
14(2884)
728
1099
8235
9913
9915
722
1168
8435
9987
9806
75053
1270
8567
9997
9834
15(3090)
855
1089
8187
9922
9861
880
1157
8376
9976
9796
89222
1300
8611
9996
9798
16(3296)
1009
1158
8770
9897
9788
966
1231
8845
9984
9823
106288
1359
8997
9999
9848
17(34102)
1154
1299
8434
9925
9849
1137
1377
8601
9980
9798
123869
1493
8798
9996
9877
18(36108)
1407
1411
8321
9927
9866
1339
1495
8301
9965
9822
133807
1601
8567
9999
9834
19(38114
)1702
1506
8206
9915
9805
1672
1598
8977
9978
9845
152999
1694
8998
9999
9847
20(40120)
2091
1635
8528
9918
9889
1916
1737
8805
9976
9837
1890
021833
9012
9999
9791
21(42126)
2843
1622
8550
9934
9913
2432
1718
8886
9974
9813
2399
131815
9008
10000
9815
22(44132)
3300
1736
8449
9940
9859
2795
1843
8786
9977
9756
253106
1948
8997
9999
9826
23(46138)
3741
1934
8133
9911
9876
3143
2050
8876
9981
9785
274583
2173
9014
9999
9858
24(48144)
4531
1961
8245
9909
9846
3515
2090
8344
9993
9819
308089
2211
8661
9999
9883
25(50150)
4715
1924
8111
9929
9888
3959
2041
8279
9996
9837
3510
932158
8527
9999
9827
26(52156)
5223
1911
8096
9898
9857
4301
2025
8251
9997
9757
421747
2146
8497
9999
9914
27(54162)
5994
2095
8387
9908
9775
4835
2226
8497
9974
9780
4614
432359
8695
9999
9837
28(56168)
6406
2111
8411
9906
9798
5336
2240
8558
9991
9820
494113
2371
9249
9999
9816
29(58174)
6897
2176
8012
9916
9791
5948
2311
8211
9984
9765
624835
2445
8531
10000
9851
30(60180)
7436
2391
8111
9913
9766
6558
2533
8305
9968
9795
696116
2682
8575
9999
9799
31(62186)
8118
2282
8119
9928
9806
7069
2422
8417
9973
9873
797768
2564
8667
9999
9846
32(64192)
8895
2519
8223
9919
9757
7857
2675
8631
9982
9769
864813
2827
8976
10000
9905
33(66198)
11235
2533
8563
9928
9811
9162
2688
8547
9979
9828
973089
2843
8657
9999
9846
34(68204)
11587
2534
8249
9926
9847
9549
2686
8798
9938
9856
10523
3041
8999
10000
9877
35(70210)
1164
92715
8098
9895
9872
10345
2880
8602
9998
9869
11272
3052
9021
9999
9897
36(72216)
12224
2825
8403
9958
9770
11165
3002
8657
9999
9839
12456
3175
9113
9999
9864
37(74222)
12978
2830
8357
9937
9795
11850
3005
8496
9998
9894
13458
3176
8827
9999
9905
38(76228)
13426
2844
8267
9966
9759
12795
3019
8604
9999
9797
15155
3121
8978
9999
9853
39(78234)
15003
2955
8303
9956
9867
13816
3132
8588
9996
9885
15041
3310
9016
9999
9912
40(80240)
16495
2978
8257
9954
9826
14896
3160
8495
9997
9837
16563
3349
8957
9999
9933
Average
4619
15855
8174
9925
9742
4077
16828
8457
9981
9743
4350
17938
8732
9999
9794
Journal of Advanced Transportation 11
high service level requirement we recommend MIP-DF1 forsolutions
6 Conclusion
This paper investigates an ambulance location problem withambiguity set of demand in which the probability distribu-tion of the uncertain demand is unknown In such a data-driven environment only the mean and covariance demandsat emergency points are knownThe problem is to determinethe base locations and the number of ambulances at eachbase aiming at minimizing the total cost associated withlocating bases and assigning ambulances We propose adistribution-free model with chance constraints Then twoapproximated MIP formulations with different approxima-tion methods of chance constraints are proposed which arebased on different ambiguity sets of the unknown probabilitydistribution Numerical experiments on benchmarks and120 randomly generated instances are conducted and thecomputational results show that the first approximated MIPformulation is much conservative with overhigh system costwhile the second approximatedMIP formulation ismore costsaving with service level appropriately ensured
In future research wemay consider the following relevantissues First the uncertainty of driving time from a base toan emergency point shall be analysed Besides unexpected orsudden disasters may be taken into consideration Moreoverother uncertainties should be considered such as travel timethe number of available vehicles and so on The stochasticdependence should also further considered
Data Availability
The data of the benchmark instances in our manuscript canbe obtained in Nickel et al [8] For the test instances thedata is randomly generated and the way for generating data isstated in our manuscript The datasets generated during thecurrent study are available from the corresponding author onreasonable request
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This work was supported by the National Natural Sci-ence Foundation of China (NSFC) under Grants 7142800271531011 71571134 and 71771048 This work was also sup-ported by the Fundamental Research Funds for the CentralUniversities
References
[1] L Brotcorne G Laporte and F Semet ldquoAmbulance loca-tion and relocation modelsrdquo European Journal of OperationalResearch vol 147 no 3 pp 451ndash463 2003
[2] X Li Z Zhao X Zhu and T Wyatt ldquoCovering modelsand optimization techniques for emergency response facilitylocation and planning a reviewrdquo Mathematical Methods ofOperations Research vol 74 no 3 pp 281ndash310 2011
[3] C Toregas R Swain C ReVelle and L Bergman ldquoThe locationof emergency service facilitiesrdquoOperations Research vol 19 no6 pp 1363ndash1373 1971
[4] R Church and C ReVelle ldquoThe maximal covering locationproblemrdquo Papers in Regional Science vol 32 no 1 pp 101ndash1181974
[5] M Gendreau G Laporte and F Semet ldquoSolving an ambulancelocation model by tabu searchrdquo Location Science vol 5 no 2pp 75ndash88 1997
[6] A Ingolfsson S Budge and E Erkut ldquoOptimal ambulancelocation with random delays and travel timesrdquo Health CareManagement Science vol 11 no 3 pp 262ndash274 2008
[7] M Gendreau G Laporte and F Semet ldquoThemaximal expectedcoverage relocation problem for emergency vehiclesrdquo Journal ofthe Operational Research Society vol 57 no 1 pp 22ndash28 2006
[8] S Nickel M Reuter-Oppermann and F Saldanha-da-GamaldquoAmbulance location under stochastic demand A samplingapproachrdquo Operations Research for Health Care vol 8 pp 24ndash32 2016
[9] P Beraldi M E Bruni and D Conforti ldquoDesigning robustemergency medical service via stochastic programmingrdquo Euro-pean Journal of Operational Research vol 158 no 1 pp 183ndash1932004
[10] N Noyan ldquoAlternate risk measures for emergency medicalservice system designrdquo Annals of Operations Research vol 181pp 559ndash589 2010
[11] M R Wagner ldquoStochastic 0-1 linear programming underlimited distributional informationrdquoOperations Research Lettersvol 36 no 2 pp 150ndash156 2008
[12] EDelage andYYe ldquoDistributionally robust optimization undermoment uncertainty with application to data-driven problemsrdquoOperations Research vol 58 no 3 pp 595ndash612 2010
[13] J T Essen J L Hurink S Nickel and M Reuter ldquoModels forambulance planning on the strategic and the tactical levelrdquo BetaResearch School for Operations Management and Logistics p 272013
[14] E Erkut A Ingolfsson and G Erdogan ldquoAmbulance locationfor maximum survivalrdquoNaval Research Logistics (NRL) vol 55no 1 pp 42ndash58 2008
[15] S C Chapman and J A White ldquoProbabilistic formulations ofemergency service facilities location problemsrdquo in Proceedingsof ORSATIMS Conference San Juan Puerto Rico 1974
[16] A Ben-Tal D Den Hertog A De Waegenaere B Melenbergand G Rennen ldquoRobust solutions of optimization problemsaffected by uncertain probabilitiesrdquo Management Science vol59 no 2 pp 341ndash357 2013
[17] MNg ldquoDistribution-free vessel deployment for liner shippingrdquoEuropean Journal of Operational Research vol 238 no 3 pp858ndash862 2014
[18] M Ng ldquoContainer vessel fleet deployment for liner shippingwith stochastic dependencies in shipping demandrdquo Transporta-tion Research Part B Methodological vol 74 pp 79ndash87 2015
[19] R Jiang and Y Guan ldquoData-driven chance constrained stochas-tic programrdquo Mathematical Programming vol 158 no 1-2 SerA pp 291ndash327 2016
[20] C-M Lee and S-L Hsu ldquoThe effect of advertising on thedistribution-free newsboy problemrdquo International Journal ofProduction Economics vol 129 no 1 pp 217ndash224 2011
[21] K Kwon and T Cheong ldquoA minimax distribution-free proce-dure for a newsvendor problem with free shippingrdquo EuropeanJournal of Operational Research vol 232 no 1 pp 234ndash2402014
12 Journal of Advanced Transportation
[22] Y Zhang S Shen and S A Erdogan ldquoDistributionally robustappointment scheduling with moment-based ambiguity setrdquoOperations Research Letters vol 45 no 2 pp 139ndash144 2017
[23] Y Zhang S Shen and S A Erdogan ldquoSolving 0-1semidefinite programs for distributionally robust allocationof surgery blocksrdquo Social Science Electronic Publishing 2017httpsssrncomabstract=2997524
[24] F Zheng J He F Chu and M Liu ldquoA new distribution-freemodel for disassembly line balancing problem with stochastictask processing timesrdquo International Journal of ProductionResearch pp 1ndash13 2018
[25] J Cheng E Delage and A Lisser ldquoDistributionally robuststochastic knapsack problemrdquo SIAM Journal on Optimizationvol 24 no 3 pp 1485ndash1506 2014
[26] M Liu D Yang and F Hao ldquoOptimization for the locationsof ambulances under two-stage life rescue in the emergencymedical service a case study in Shanghai ChinardquoMathematicalProblems in Engineering vol 2017 Article ID 1830480 14 pages2017
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
6 Journal of Advanced Transportation
Table 2 120572 = 099|119868| = |119869| Sampling approach MIP-DF1 MIP-DF2
CT Obj Sel () CT Obj Sel () CT Obj Sel ()1 7 49 24 9979 01 64 10000 01 25 100002 7 48 23 9856 01 63 10000 01 24 100003 6 51 15 9717 01 55 10000 01 16 100004 6 47 15 9275 01 52 10000 01 13 100005 6 40 14 9000 01 55 10000 01 14 100006 8 52 23 9785 01 45 10000 01 27 10000
Average 48 190 9343 01 557 10000 01 198 10000
Table 3 Solutions to instance 4 in Nickel et al [8] for the given scenario
Sampling approach MIP-DF1 MIP-DF2Emergency points 1 2 3 4 5 6 Emergency points 1 2 3 4 5 6 Emergency points 1 2 3 4 5 6
Demand 1 0 2 0 2 0 Demand 1 0 2 0 2 0 Demand 1 0 2 0 2 0Base locations Ambulance employment Ambulance employment Ambulance employment1 0 0 02 1 1 0 3 1 23 1 1 26 1 2 2 04 0 0 05 0 6 06 2 2 0 4 2
The average service level is 9343 Although the averageservice levels obtained by MIP-DF1 and MIP-DF2 are both100 the average system cost obtained by MIP-DF1 is about3 times larger than that of sampling approach Besides theaverage system cost obtained by MIP-DF2 is 421 higherthan that of sampling approach
Concluding for the benchmarks described in Nickel etal [8] we can observe that (1) MIP-DF1 and MIP-DF2 canbe solved in far less computational time compared with thesampling approach (2) the sampling approach cannot meetthe requirement of service level for all benchmarks whilethe two MIP formulations can (3) the average system cost ofMIP-DF2 is very close to that of sampling approach (4) withthe increase of 120572 the average system costs obtained by MIP-DF2 is getting closer to that obtained by sampling approach
From Tables 1 and 2 we can observe that in spite ofthe lower system cost obtained by the sampling approachthere may exist extreme situations that demands of someemergency points are not satisfied Take one benchmark inNickel et al [8] ie instance 4 where there are 6 nodesA scenario in which demand at each emergency point isgenerated from the given probability distribution is shown inFigure 1 We can observe from Figure 1 that under the givenscenario the demand of each possible emergency point is1 0 2 0 2 0The solutions obtained by sampling approachMIP-DF1 andMIP-DF2 are shown in Table 3 which includesbase locations and ambulance employment obtained by in-sample test and ambulance assignment under given scenario
In Table 3 the first column includes the indexes of poten-tial base locations The second column represents the baselocations and the ambulance employment at each location
>1 = 1
>2 = 0
>3 = 2
>4 = 0
>5 = 2
>6 = 0
1
2 4
3 5
6
Figure 1 Instance 4 in Nickel et al [8] under given scenario
in which a location with 0 ambulance implies that it is notselected for base construction From Table 3 we can observethat the in-sample solutions obtained by sampling approachis to construct bases at locations 2 3 and 6 where the numberof ambulances is 1 1 and 2 respectively Besides undergiven scenario there is one ambulance serving emergencypoint 1 from base location 3 and one ambulance servingemergency point from base location 2 and 2 ambulancesserving emergency point 5 from base location 6 It shows thatthe requirement of ambulances at emergency point 3 is notfully satisfied The insufficient service may cause serious lossof life However we can find that the solutions obtained byMIP-DF1 andMIP-DF2 can provide sufficient ambulances forall emergency points
In sum to guarantee the required service level withless cost MIP-DF2 is recommended for the solutions with
Journal of Advanced Transportation 7
Nickel et al [8]rsquos data set Moreover in the benchmarks itis assumed that all nodes are possible emergency points aswell as potential base locations and the number of requiredambulances is small which is not always consistent withthe situation in Shanghai Therefore we further test the twoapproximated MIP formulations and the sampling approachon randomly generated instances by analysing the informa-tion on emergencies in Shanghai [26]
53 Computational Tests on Randomly Generated InstancesIn the following subsection numerical experiments on 120randomly generated instances are described
531 Test Instances By analysing the information on emer-gencies in Shanghai [26] the mean value 120583119895 of demand 119889119895 ineach emergency point are randomly generated fromadiscreteUniform distribution on interval [5 25] The standard devia-tion 120590119895 is set to be 5 Since the sampling approach proposed inNickel et al [8] requires a sample of scenarios in numericalexperiments scenarios are produced with demands randomlygenerated from Uniform distribution Poisson distributionand Normal distribution consistent with given mean andstandard deviation
According to the instance settings in Erkut and Ingolfsson(2008) response time from a base to an emergency pointis randomly generated from a discrete Uniform distributionon interval [3 30] in units of minutes A base covers anemergency point if the driving time between them is lessthan 10 minutes to ensure the survival probability (Erkut andIngolfsson 2006) For an emergency point in each instanceif there is no base with 10 minutes driving time to it then thebase covering this point is set to be the nearest one The costof constructing a base and that of locating one ambulance areboth set to be 1 as in Nickel et al [8] Besides the number ofemergency points is 3 times that of potential base locations
532 Results and Discussion Different values of requiredservice level 120572 are tested ie 085 09 095 Computationalresults on 120 randomly generated instances are reported inTables 4-6
Table 4 reports the numerical results of instances withdemand generated from Uniform distribution We observethat when 120572 = 085 09 095 the average computationaltimes of the sampling approach are 4482 4379 and 4295seconds respectively However both approximated MIP for-mulations can produce solutions in much faster speeds iewithin 13 seconds on average With the increase of 120572 (i)the out-of-sample average service level is getting higher and(ii) the system costs of solutions obtained by all method aregetting larger and (iii) the system costs of solutions yieldedbyMIP-DF2 is getting closer to those of solutions obtained bythe sampling approach In terms of the average system costthe sampling approach performs the best while MIP-DF1 isthe poorest When 120572 = 085 for example the average systemcost ofMIP-DF2 and the sampling approach are about 3032and 442 less than that of MIP-DF1 respectively When120572 = 095 the above gaps are even larger Concluding fortheUniformdistribution (1) the proposed two approximatedMIP formulations are far more time saving than the sampling
approach (2)MIP-DF1 andMIP-DF2 can improve the servicelevel by about 1907 compared with the sampling approachhowever (3) despite of the highest service levels obtainedby MIP-DF1 the system cost obtained by MIP-DF1 is alsohigh (4) MIP-DF2 can ensure an appropriate service levelwith much less cost than MIP-DF1 and (5) for the situationswith small-scale instances and high service level requirementMIP-DF1 is recommended
Computational results of instances with demand at eachemergency point generated from Poisson distribution arepresented in Table 5 For MIP-DF1 and MIP-DF2 onlymean and variance are involved Therefore with givenmean and standard deviation of the demand at each pointcomputational results are the same for different probabilitydistributions We can obtain from Table 5 that the objectivevalue obtained by the sampling approach increases with theincrease of 120572 Besides when the value of 120572 equals 085the average service level obtained by the sampling approachis about 8143 which is smaller than that of the twoapproximated MIP formulations The highest service level9934 is obtained by MIP-DF1 and that obtained by MIP-DF2 is 9708 When 120572 = 09 the average service levelof the sampling approach is 8437 and those obtained byMIP-DF1 and MIP-DF2 are 9985 and 9797 respectivelyWhen 120572 = 095 the average service level obtained by thesampling approach is 9356 and those yielded by MIP-DF1 and MIP-DF2 are 9997 and 9781 respectively ForPoisson distribution concluding MIP-DF1 and MIP-DF2can improve the service level compared with the samplingapproach especially MIP-DF1 can obtain the highest servicelevel
Similarly Table 6 reports the computational results forinstances in which the demand at each emergency point isgenerated following Normal distribution When 120572 = 085the average service levels obtained by the sampling approachMIP-DF1 and MIP-DF2 are 8174 9925 and 9742respectively When 120572 = 09 the three values are 84579981 and 9743 respectively Moreover for 120572 = 095 thevalues are equal to 8732 9999 and 9794 respectivelyIt can be observed that the performance of the samplingapproach is the poorest in terms of the service level
By observing the above computational results on bench-marks and randomly generated instances we conclude that(1)The proposed approximated MIP formulations ie MIP-DF1 and MIP-DF2 are very time saving compared with thesampling approach (2) MIP-DF1 an MIP-DF2 can achievethe requirement of service level and improve the service levelby about 1547 and 1229 on average compared with thatof the sampling approach (3) MIP-DF1 is more conservativein terms of overhigh emergency service level close to 100and the system cost is also high which is about 568 largerthan that of MIP-DF2 (4) Compared with MIP-DF1 MIP-DF2 is less conservative and much more cost saving whileensuring an appropriately service level on average which isabout 97
In sum for medium and large problem instances (egShanghai) which are common in practice to guarantee therequired service level with less cost we recommendMIP-DF2as solution method However for very small instances with
8 Journal of Advanced Transportation
Table4Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mun
iform
distrib
ution
Set (|119868||119869|)
120572=085
120572=09
120572=095
Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
1(26)
27
897317
01
149
9906
01
107
9222
28
967373
01
155
9933
01
107
9167
27
106
7585
01
185
9987
01
107
9128
2(412
)47
169
7423
00
291
9919
00
207
9122
58
181
7438
01
303
9997
01
207
9363
50
199
7702
00
363
9982
00
207
9338
3(618
)72
258
7506
01
438
9976
01
312
9106
75279
8082
01
456
9996
01
312
9369
78304
8337
01
546
9997
01
312
9357
4(824)
99314
7956
01
553
9993
01
385
9206
111
338
8202
01
577
9999
01
385
9602
112
366
8492
01
697
9994
01
385
9620
5(1030)
132
415
8406
01
722
9896
01
512
9406
143
441
8322
02
752
9983
01
512
9692
142
476
8609
01
902
9993
01
512
9692
6(1236)
170
475
8249
02
843
9870
01
591
9367
181
504
8251
01
879
9982
01
591
9608
189
535
8726
01
1059
9997
01
591
9609
7(14
42)
215
573
8111
02
1006
9849
02
712
9606
243
609
8217
02
1048
9953
02
712
9589
233
644
8608
02
1258
9992
02
712
9582
8(1648)
248
686
7532
02
1185
9896
02
849
9725
284
727
8307
02
1233
9949
02
849
9613
281
768
9063
02
1473
9999
03
849
9622
9(1854)
313
691
8353
03
1236
9974
04
858
9709
351
731
8397
03
1290
9988
03
858
9765
352
760
8906
03
1560
9994
03
858
9758
10(2060)
372
794
8174
03
1409
9887
03
989
9657
428
843
8489
03
1469
9947
03
989
9773
432
892
9004
03
1769
9997
03
989
9738
11(2266
)451
830
8072
04
1493
9949
03
1031
9758
518
880
8328
04
1559
9982
03
1031
9723
516
930
8967
05
1889
9995
03
1031
9727
12(2472)
543
1025
7970
04
1771
9927
04
1267
9835
601
1083
8482
05
1843
9949
04
1267
9776
637
1146
9014
04
2203
9993
04
1267
9777
13(2678)
631
1026
8142
05
1819
9898
04
1274
9782
743
1087
8678
05
1897
9936
04
1274
9778
765
1146
9023
04
2287
9997
05
1273
9837
14(2884)
782
1101
7903
05
1949
9869
08
1361
9804
878
1165
8790
05
2033
9935
07
1361
9889
870
1227
9056
05
2453
9998
06
1361
9808
15(3090)
907
1091
8098
06
1997
9915
06
1367
9629
1018
1155
8329
06
2087
9981
06
1367
9807
1020
1225
8986
06
2537
9999
08
1367
9817
16(3296)
1062
1160
8798
07
2123
9944
07
1449
9787
1183
1229
8383
07
2219
9988
06
1449
9849
1193
1300
9105
06
2698
9998
06
1449
9815
17(34102)
1202
1298
7572
07
2328
9969
07
1614
9821
1377
1374
8860
08
2430
9997
07
1614
9824
1260
1452
9085
07
2940
9999
08
1614
9818
18(36108)
1394
1408
8121
08
2509
9947
08
1752
9809
1544
1492
8928
08
2617
9983
08
1752
9799
1548
1580
9173
08
3156
9999
08
1752
9809
19(38114
)1618
1508
8170
09
2676
9899
111877
9742
1760
1598
9130
09
2790
9934
08
1877
9908
1755
1688
9186
09
3360
9999
09
1877
9872
20(40120)
2206
1628
7719
102869
9989
102029
9873
2007
1732
8753
112990
9997
09
2029
9879
1904
1827
9138
103590
9998
09
2029
9875
21(42126)
2780
1622
8354
102900
9978
102017
9732
2609
1718
8854
103026
9876
102017
9818
2670
1812
9089
123655
9999
102017
9848
22(44132)
3189
1735
8494
113088
9985
152165
9857
2970
1840
8922
123220
9925
122165
9886
2873
1952
9206
113880
9999
112165
9850
23(46138)
3630
1937
8141
123372
9893
132406
9796
3222
2048
8937
123510
9934
142406
9906
3461
2170
8989
124199
10000
122406
9809
24(48144)
4226
1968
8226
133461
9932
132453
9814
3674
2085
8846
133605
9956
152453
9912
3556
2211
9300
134326
9999
132453
9819
25(50150)
4727
1919
8111
143456
9999
152407
9629
4019
2037
8699
143606
9999
142407
9857
4257
2161
9259
154356
9995
142407
9819
26(52156)
5300
1906
8096
153484
9878
172392
9658
4495
2021
8952
143640
9999
152392
9568
4486
2142
9182
1544
209999
152392
9854
27(54162)
5913
2096
8387
163751
9846
172617
9761
5117
2221
9045
163913
9998
172617
9544
4830
2355
9346
184723
9999
162617
9893
28(56168)
6310
2110
7758
173819
9933
192643
9829
5777
2234
9157
183987
9998
182643
9526
5550
2365
9045
184827
9999
172643
9831
29(58174)
6780
2175
7998
183937
9898
182720
9714
6303
2307
9022
20
4111
9969
192720
9623
6189
2444
9128
194981
10000
192720
9914
30(60180)
7593
2384
8321
194240
9967
20
2979
9841
6791
2527
8978
1944
199979
23
2979
9721
6878
2679
9143
21
5316
9999
20
2979
9869
31(62186)
8220
2281
7604
20
4157
9984
21
2856
9796
7734
2416
8934
21
4343
9997
21
2856
9679
7466
2557
9158
22
5273
9999
23
2856
9902
32(64192)
8766
2517
8378
21
4489
9932
25
3145
9729
8276
2671
8296
23
4680
9977
24
3145
9825
7897
2822
9089
24
5640
9999
22
3145
9891
33(66198)
10123
2532
7806
23
4555
9967
23
3169
9615
9010
2682
8947
24
4752
9989
24
3169
9703
9051
2835
9157
24
5742
9998
24
3169
9867
34(68204)
11010
2535
7934
24
4602
9999
29
3174
9631
9898
2685
8858
30
4806
9998
28
3174
9762
9863
2844
9236
26
5824
9999
25
3174
9889
35(70210)
11666
2714
8062
25
4862
9986
29
3391
9787
1066
62875
9137
26
5072
9999
29
3391
9716
10822
3043
9315
26
6122
9999
29
3391
9923
36(72216)
12011
2827
8104
26
5047
9991
28
3535
9798
1246
52984
9016
28
5263
9999
28
3535
9878
11667
3169
9253
30
6343
9999
29
3535
9896
37(74222)
11437
2826
8228
28
5096
9933
29
3543
9884
13582
3098
8895
29
5318
9999
32
3543
9888
12654
3178
9139
29
6427
10000
31
3543
9893
38(76228)
12435
2839
8152
29
5161
9897
29
3565
9788
13976
3103
8926
29
5388
9999
31
3565
9724
13837
3185
9255
33
6528
10000
31
3565
9895
39(78234)
1464
32945
8412
31
5331
9998
32
3695
9876
14794
3210
9005
31
5565
9999
34
3695
9713
14561
3313
9081
34
6735
10000
33
3695
9878
40(80240)
1604
22975
8510
32
5414
9957
36
3734
9796
16263
3346
8999
35
5654
9999
35
3734
9806
15869
3343
9248
35
6854
10000
35
3734
9873
Average
4482
1584
680
67
1228
397
9936
131978
796
82
4379
16913
8654
1329
626
9975
131978
79721
4295
1778
889
85
1335774
9997
131978
79775
Journal of Advanced Transportation 9
Table5Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mPo
isson
distr
ibution
set ( |119868 |
|119869 |)120572=
085120572=
09120572=
095Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
1(26)
27
907294
9967
9001
27
987770
9979
9013
24
103
9273
9999
9468
2(412
)46
170
7961
9921
9117
51
183
8305
9946
9128
49
201
9392
9998
9546
3(618
)68
258
7578
9938
8911
74280
8028
9968
9004
69
307
9182
9999
9658
4(824)
93315
8001
9928
9196
102
338
8438
9967
9172
92368
9468
9999
9613
5(1030)
120
413
8041
9939
9449
133
443
8447
9976
9272
121
479
9544
9999
9752
6(1236)
155
476
8084
9877
9554
167
508
8863
9991
9569
159
536
9456
9999
9734
7(14
42)
195
573
7876
9892
9663
213
609
9124
9998
9729
205
643
9314
10000
9768
8(1648)
232
687
7964
9997
9788
267
724
8523
9988
9687
238
772
9422
9998
9689
9(1854)
293
692
8267
9893
9809
327
730
8386
9999
9750
298
757
9228
9999
9755
10(2060)
356
796
7946
9899
9801
399
840
8299
9976
9686
360
885
9668
9999
9705
11(2266)
423
829
8201
9895
9756
489
877
8378
9987
9788
448
927
9476
9999
9812
12(2472)
512
1023
8338
9913
9853
554
1085
8501
9999
9881
514
1148
9398
9999
9888
13(2678)
592
1045
8416
9907
9873
687
1088
7998
9996
9802
607
1152
9504
9999
9821
14(2884)
723
1099
7946
9915
9788
838
1166
8056
9992
9829
776
1230
9546
9999
9839
15(3090)
855
1089
7915
9964
9703
942
1156
8432
9984
9778
903
1225
9223
9998
9785
16(3296)
1009
1158
8024
9897
9769
1163
1231
8500
9996
9824
1045
1222
9199
9999
9835
17(34102)
1165
1299
8034
9919
9779
1285
1375
8495
9937
9808
1190
1457
9258
10000
9846
18(36108)
1254
1411
8176
9943
9885
1479
1492
8561
9968
9868
1345
1579
9286
9999
9875
19(38114
)1504
1506
7970
9923
9842
1651
1594
8167
9972
9846
1564
1694
9376
9999
9786
20(40120)
1706
1635
7819
9916
9783
1854
1734
8607
9968
9776
1778
1825
9285
9904
9718
21(42126)
2054
1622
8354
9918
9834
2357
1720
8297
9978
9833
2226
1856
9501
9999
9834
22(44132)
2400
1736
8494
9922
9846
2852
1844
8269
9999
9816
2680
1946
9486
9999
9820
23(46138)
2651
1934
8141
9953
9793
3159
2049
8492
9969
9799
3072
2172
9358
10000
9813
24(48144)
3004
1961
8226
9947
9681
3522
2086
8513
9998
9706
3102
2209
9488
9999
9724
25(50150)
3308
1924
8111
9965
9746
3898
2038
8264
9999
9788
3639
2161
9546
9999
9813
26(52156)
3669
1911
8096
9939
9855
4305
2027
8722
9967
9874
4453
2142
9185
9999
9876
27(54162)
4418
2095
8387
9927
9754
4771
2219
8341
9989
9749
4806
2353
9289
9999
9755
28(56168)
5367
2113
8458
9968
9848
5356
2237
8452
9978
9855
5221
2369
9228
10000
9868
29(58174)
6217
2186
7998
9989
9694
5942
2315
8316
9996
9699
5853
2371
9167
9999
9705
30(60180)
7263
2382
8321
9991
9837
6527
2543
8359
9998
9886
6214
2678
9369
9997
9886
31(62186)
8214
2346
8404
9936
9888
7813
2430
8321
9999
9846
6876
2560
9358
9999
9847
32(64192)
8863
2518
8378
9938
9807
8106
2677
8356
9999
9833
7877
2822
9408
10000
9835
33(66198)
9985
2532
8406
9917
9709
9684
2689
8497
9996
9746
8088
2838
9507
9999
9749
34(68204)
11264
2534
8543
9908
9897
10986
2685
8809
9996
9782
8469
2843
8844
9998
9780
35(70210)
12035
2715
8258
9937
9799
12098
2886
8569
9991
9809
9946
3041
9459
9999
9813
36(72216)
12803
2825
8104
9929
9827
13076
3010
8651
9994
9834
11013
3170
9468
9999
9865
37(74222)
13021
2830
8228
9949
9826
13261
3009
8376
9999
9826
12014
3172
9346
9999
9888
38(76228)
13862
2843
8152
9933
9807
13995
3028
8760
9998
9841
12133
3187
9706
9999
9876
39(78234)
15387
2953
8312
9964
9799
14827
3133
9001
9995
9783
13381
3304
8913
9999
9799
40(80240)
16974
2975
8510
9985
9767
15256
3249
8217
9993
9781
15867
3347
9111
9999
9791
Average
4352
15875
8143
9934
9708
4362
16856
8437
9985
9707
3968
17763
9356
9997
9781
10 Journal of Advanced Transportation
Table6Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mno
rmaldistr
ibution
set ( |119868 |
|119869 |)120572=
085120572=
09120572=
095Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
1(26)
27
907294
9964
9117
27
947844
9970
9202
2558
106
8422
10000
9208
2(412
)46
170
7961
9921
9283
48
183
8328
9982
9283
4508
183
8823
10000
9370
3(618
)68
258
7578
9916
9228
68
280
7798
9971
9417
6803
307
7658
9998
9447
4(824)
93315
8000
9899
9367
96337
8379
9979
9467
9373
368
8438
9999
9579
5(1030)
120
413
7841
9927
9437
124
443
8763
9963
9476
1262
499
8850
9998
9708
6(1236)
155
476
7921
9919
9519
162
504
8124
9982
9582
16021
536
8954
10000
9756
7(14
42)
195
573
7976
9933
9598
196
609
8287
9975
9619
2011
646
8467
9999
9767
8(1648)
232
687
8031
9918
9705
235
727
8115
9965
9650
2404
772
8278
9999
9765
9(1854)
293
692
8086
9898
9759
291
734
8464
9974
9755
29815
796
8617
9998
9835
10(2060)
356
796
7984
9937
9770
358
842
8223
9988
9765
37973
921
8846
10000
9838
11(2266
)421
829
8101
9940
9755
436
877
8211
9986
9786
4514
8959
8497
9999
9789
12(2472)
514
1023
8033
9951
9785
501
1085
8162
9983
9788
52073
1147
8321
9999
9824
13(2678)
592
1023
8138
9916
9885
599
1087
8411
9970
9845
62937
1194
8609
9998
9795
14(2884)
728
1099
8235
9913
9915
722
1168
8435
9987
9806
75053
1270
8567
9997
9834
15(3090)
855
1089
8187
9922
9861
880
1157
8376
9976
9796
89222
1300
8611
9996
9798
16(3296)
1009
1158
8770
9897
9788
966
1231
8845
9984
9823
106288
1359
8997
9999
9848
17(34102)
1154
1299
8434
9925
9849
1137
1377
8601
9980
9798
123869
1493
8798
9996
9877
18(36108)
1407
1411
8321
9927
9866
1339
1495
8301
9965
9822
133807
1601
8567
9999
9834
19(38114
)1702
1506
8206
9915
9805
1672
1598
8977
9978
9845
152999
1694
8998
9999
9847
20(40120)
2091
1635
8528
9918
9889
1916
1737
8805
9976
9837
1890
021833
9012
9999
9791
21(42126)
2843
1622
8550
9934
9913
2432
1718
8886
9974
9813
2399
131815
9008
10000
9815
22(44132)
3300
1736
8449
9940
9859
2795
1843
8786
9977
9756
253106
1948
8997
9999
9826
23(46138)
3741
1934
8133
9911
9876
3143
2050
8876
9981
9785
274583
2173
9014
9999
9858
24(48144)
4531
1961
8245
9909
9846
3515
2090
8344
9993
9819
308089
2211
8661
9999
9883
25(50150)
4715
1924
8111
9929
9888
3959
2041
8279
9996
9837
3510
932158
8527
9999
9827
26(52156)
5223
1911
8096
9898
9857
4301
2025
8251
9997
9757
421747
2146
8497
9999
9914
27(54162)
5994
2095
8387
9908
9775
4835
2226
8497
9974
9780
4614
432359
8695
9999
9837
28(56168)
6406
2111
8411
9906
9798
5336
2240
8558
9991
9820
494113
2371
9249
9999
9816
29(58174)
6897
2176
8012
9916
9791
5948
2311
8211
9984
9765
624835
2445
8531
10000
9851
30(60180)
7436
2391
8111
9913
9766
6558
2533
8305
9968
9795
696116
2682
8575
9999
9799
31(62186)
8118
2282
8119
9928
9806
7069
2422
8417
9973
9873
797768
2564
8667
9999
9846
32(64192)
8895
2519
8223
9919
9757
7857
2675
8631
9982
9769
864813
2827
8976
10000
9905
33(66198)
11235
2533
8563
9928
9811
9162
2688
8547
9979
9828
973089
2843
8657
9999
9846
34(68204)
11587
2534
8249
9926
9847
9549
2686
8798
9938
9856
10523
3041
8999
10000
9877
35(70210)
1164
92715
8098
9895
9872
10345
2880
8602
9998
9869
11272
3052
9021
9999
9897
36(72216)
12224
2825
8403
9958
9770
11165
3002
8657
9999
9839
12456
3175
9113
9999
9864
37(74222)
12978
2830
8357
9937
9795
11850
3005
8496
9998
9894
13458
3176
8827
9999
9905
38(76228)
13426
2844
8267
9966
9759
12795
3019
8604
9999
9797
15155
3121
8978
9999
9853
39(78234)
15003
2955
8303
9956
9867
13816
3132
8588
9996
9885
15041
3310
9016
9999
9912
40(80240)
16495
2978
8257
9954
9826
14896
3160
8495
9997
9837
16563
3349
8957
9999
9933
Average
4619
15855
8174
9925
9742
4077
16828
8457
9981
9743
4350
17938
8732
9999
9794
Journal of Advanced Transportation 11
high service level requirement we recommend MIP-DF1 forsolutions
6 Conclusion
This paper investigates an ambulance location problem withambiguity set of demand in which the probability distribu-tion of the uncertain demand is unknown In such a data-driven environment only the mean and covariance demandsat emergency points are knownThe problem is to determinethe base locations and the number of ambulances at eachbase aiming at minimizing the total cost associated withlocating bases and assigning ambulances We propose adistribution-free model with chance constraints Then twoapproximated MIP formulations with different approxima-tion methods of chance constraints are proposed which arebased on different ambiguity sets of the unknown probabilitydistribution Numerical experiments on benchmarks and120 randomly generated instances are conducted and thecomputational results show that the first approximated MIPformulation is much conservative with overhigh system costwhile the second approximatedMIP formulation ismore costsaving with service level appropriately ensured
In future research wemay consider the following relevantissues First the uncertainty of driving time from a base toan emergency point shall be analysed Besides unexpected orsudden disasters may be taken into consideration Moreoverother uncertainties should be considered such as travel timethe number of available vehicles and so on The stochasticdependence should also further considered
Data Availability
The data of the benchmark instances in our manuscript canbe obtained in Nickel et al [8] For the test instances thedata is randomly generated and the way for generating data isstated in our manuscript The datasets generated during thecurrent study are available from the corresponding author onreasonable request
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This work was supported by the National Natural Sci-ence Foundation of China (NSFC) under Grants 7142800271531011 71571134 and 71771048 This work was also sup-ported by the Fundamental Research Funds for the CentralUniversities
References
[1] L Brotcorne G Laporte and F Semet ldquoAmbulance loca-tion and relocation modelsrdquo European Journal of OperationalResearch vol 147 no 3 pp 451ndash463 2003
[2] X Li Z Zhao X Zhu and T Wyatt ldquoCovering modelsand optimization techniques for emergency response facilitylocation and planning a reviewrdquo Mathematical Methods ofOperations Research vol 74 no 3 pp 281ndash310 2011
[3] C Toregas R Swain C ReVelle and L Bergman ldquoThe locationof emergency service facilitiesrdquoOperations Research vol 19 no6 pp 1363ndash1373 1971
[4] R Church and C ReVelle ldquoThe maximal covering locationproblemrdquo Papers in Regional Science vol 32 no 1 pp 101ndash1181974
[5] M Gendreau G Laporte and F Semet ldquoSolving an ambulancelocation model by tabu searchrdquo Location Science vol 5 no 2pp 75ndash88 1997
[6] A Ingolfsson S Budge and E Erkut ldquoOptimal ambulancelocation with random delays and travel timesrdquo Health CareManagement Science vol 11 no 3 pp 262ndash274 2008
[7] M Gendreau G Laporte and F Semet ldquoThemaximal expectedcoverage relocation problem for emergency vehiclesrdquo Journal ofthe Operational Research Society vol 57 no 1 pp 22ndash28 2006
[8] S Nickel M Reuter-Oppermann and F Saldanha-da-GamaldquoAmbulance location under stochastic demand A samplingapproachrdquo Operations Research for Health Care vol 8 pp 24ndash32 2016
[9] P Beraldi M E Bruni and D Conforti ldquoDesigning robustemergency medical service via stochastic programmingrdquo Euro-pean Journal of Operational Research vol 158 no 1 pp 183ndash1932004
[10] N Noyan ldquoAlternate risk measures for emergency medicalservice system designrdquo Annals of Operations Research vol 181pp 559ndash589 2010
[11] M R Wagner ldquoStochastic 0-1 linear programming underlimited distributional informationrdquoOperations Research Lettersvol 36 no 2 pp 150ndash156 2008
[12] EDelage andYYe ldquoDistributionally robust optimization undermoment uncertainty with application to data-driven problemsrdquoOperations Research vol 58 no 3 pp 595ndash612 2010
[13] J T Essen J L Hurink S Nickel and M Reuter ldquoModels forambulance planning on the strategic and the tactical levelrdquo BetaResearch School for Operations Management and Logistics p 272013
[14] E Erkut A Ingolfsson and G Erdogan ldquoAmbulance locationfor maximum survivalrdquoNaval Research Logistics (NRL) vol 55no 1 pp 42ndash58 2008
[15] S C Chapman and J A White ldquoProbabilistic formulations ofemergency service facilities location problemsrdquo in Proceedingsof ORSATIMS Conference San Juan Puerto Rico 1974
[16] A Ben-Tal D Den Hertog A De Waegenaere B Melenbergand G Rennen ldquoRobust solutions of optimization problemsaffected by uncertain probabilitiesrdquo Management Science vol59 no 2 pp 341ndash357 2013
[17] MNg ldquoDistribution-free vessel deployment for liner shippingrdquoEuropean Journal of Operational Research vol 238 no 3 pp858ndash862 2014
[18] M Ng ldquoContainer vessel fleet deployment for liner shippingwith stochastic dependencies in shipping demandrdquo Transporta-tion Research Part B Methodological vol 74 pp 79ndash87 2015
[19] R Jiang and Y Guan ldquoData-driven chance constrained stochas-tic programrdquo Mathematical Programming vol 158 no 1-2 SerA pp 291ndash327 2016
[20] C-M Lee and S-L Hsu ldquoThe effect of advertising on thedistribution-free newsboy problemrdquo International Journal ofProduction Economics vol 129 no 1 pp 217ndash224 2011
[21] K Kwon and T Cheong ldquoA minimax distribution-free proce-dure for a newsvendor problem with free shippingrdquo EuropeanJournal of Operational Research vol 232 no 1 pp 234ndash2402014
12 Journal of Advanced Transportation
[22] Y Zhang S Shen and S A Erdogan ldquoDistributionally robustappointment scheduling with moment-based ambiguity setrdquoOperations Research Letters vol 45 no 2 pp 139ndash144 2017
[23] Y Zhang S Shen and S A Erdogan ldquoSolving 0-1semidefinite programs for distributionally robust allocationof surgery blocksrdquo Social Science Electronic Publishing 2017httpsssrncomabstract=2997524
[24] F Zheng J He F Chu and M Liu ldquoA new distribution-freemodel for disassembly line balancing problem with stochastictask processing timesrdquo International Journal of ProductionResearch pp 1ndash13 2018
[25] J Cheng E Delage and A Lisser ldquoDistributionally robuststochastic knapsack problemrdquo SIAM Journal on Optimizationvol 24 no 3 pp 1485ndash1506 2014
[26] M Liu D Yang and F Hao ldquoOptimization for the locationsof ambulances under two-stage life rescue in the emergencymedical service a case study in Shanghai ChinardquoMathematicalProblems in Engineering vol 2017 Article ID 1830480 14 pages2017
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
Journal of Advanced Transportation 7
Nickel et al [8]rsquos data set Moreover in the benchmarks itis assumed that all nodes are possible emergency points aswell as potential base locations and the number of requiredambulances is small which is not always consistent withthe situation in Shanghai Therefore we further test the twoapproximated MIP formulations and the sampling approachon randomly generated instances by analysing the informa-tion on emergencies in Shanghai [26]
53 Computational Tests on Randomly Generated InstancesIn the following subsection numerical experiments on 120randomly generated instances are described
531 Test Instances By analysing the information on emer-gencies in Shanghai [26] the mean value 120583119895 of demand 119889119895 ineach emergency point are randomly generated fromadiscreteUniform distribution on interval [5 25] The standard devia-tion 120590119895 is set to be 5 Since the sampling approach proposed inNickel et al [8] requires a sample of scenarios in numericalexperiments scenarios are produced with demands randomlygenerated from Uniform distribution Poisson distributionand Normal distribution consistent with given mean andstandard deviation
According to the instance settings in Erkut and Ingolfsson(2008) response time from a base to an emergency pointis randomly generated from a discrete Uniform distributionon interval [3 30] in units of minutes A base covers anemergency point if the driving time between them is lessthan 10 minutes to ensure the survival probability (Erkut andIngolfsson 2006) For an emergency point in each instanceif there is no base with 10 minutes driving time to it then thebase covering this point is set to be the nearest one The costof constructing a base and that of locating one ambulance areboth set to be 1 as in Nickel et al [8] Besides the number ofemergency points is 3 times that of potential base locations
532 Results and Discussion Different values of requiredservice level 120572 are tested ie 085 09 095 Computationalresults on 120 randomly generated instances are reported inTables 4-6
Table 4 reports the numerical results of instances withdemand generated from Uniform distribution We observethat when 120572 = 085 09 095 the average computationaltimes of the sampling approach are 4482 4379 and 4295seconds respectively However both approximated MIP for-mulations can produce solutions in much faster speeds iewithin 13 seconds on average With the increase of 120572 (i)the out-of-sample average service level is getting higher and(ii) the system costs of solutions obtained by all method aregetting larger and (iii) the system costs of solutions yieldedbyMIP-DF2 is getting closer to those of solutions obtained bythe sampling approach In terms of the average system costthe sampling approach performs the best while MIP-DF1 isthe poorest When 120572 = 085 for example the average systemcost ofMIP-DF2 and the sampling approach are about 3032and 442 less than that of MIP-DF1 respectively When120572 = 095 the above gaps are even larger Concluding fortheUniformdistribution (1) the proposed two approximatedMIP formulations are far more time saving than the sampling
approach (2)MIP-DF1 andMIP-DF2 can improve the servicelevel by about 1907 compared with the sampling approachhowever (3) despite of the highest service levels obtainedby MIP-DF1 the system cost obtained by MIP-DF1 is alsohigh (4) MIP-DF2 can ensure an appropriate service levelwith much less cost than MIP-DF1 and (5) for the situationswith small-scale instances and high service level requirementMIP-DF1 is recommended
Computational results of instances with demand at eachemergency point generated from Poisson distribution arepresented in Table 5 For MIP-DF1 and MIP-DF2 onlymean and variance are involved Therefore with givenmean and standard deviation of the demand at each pointcomputational results are the same for different probabilitydistributions We can obtain from Table 5 that the objectivevalue obtained by the sampling approach increases with theincrease of 120572 Besides when the value of 120572 equals 085the average service level obtained by the sampling approachis about 8143 which is smaller than that of the twoapproximated MIP formulations The highest service level9934 is obtained by MIP-DF1 and that obtained by MIP-DF2 is 9708 When 120572 = 09 the average service levelof the sampling approach is 8437 and those obtained byMIP-DF1 and MIP-DF2 are 9985 and 9797 respectivelyWhen 120572 = 095 the average service level obtained by thesampling approach is 9356 and those yielded by MIP-DF1 and MIP-DF2 are 9997 and 9781 respectively ForPoisson distribution concluding MIP-DF1 and MIP-DF2can improve the service level compared with the samplingapproach especially MIP-DF1 can obtain the highest servicelevel
Similarly Table 6 reports the computational results forinstances in which the demand at each emergency point isgenerated following Normal distribution When 120572 = 085the average service levels obtained by the sampling approachMIP-DF1 and MIP-DF2 are 8174 9925 and 9742respectively When 120572 = 09 the three values are 84579981 and 9743 respectively Moreover for 120572 = 095 thevalues are equal to 8732 9999 and 9794 respectivelyIt can be observed that the performance of the samplingapproach is the poorest in terms of the service level
By observing the above computational results on bench-marks and randomly generated instances we conclude that(1)The proposed approximated MIP formulations ie MIP-DF1 and MIP-DF2 are very time saving compared with thesampling approach (2) MIP-DF1 an MIP-DF2 can achievethe requirement of service level and improve the service levelby about 1547 and 1229 on average compared with thatof the sampling approach (3) MIP-DF1 is more conservativein terms of overhigh emergency service level close to 100and the system cost is also high which is about 568 largerthan that of MIP-DF2 (4) Compared with MIP-DF1 MIP-DF2 is less conservative and much more cost saving whileensuring an appropriately service level on average which isabout 97
In sum for medium and large problem instances (egShanghai) which are common in practice to guarantee therequired service level with less cost we recommendMIP-DF2as solution method However for very small instances with
8 Journal of Advanced Transportation
Table4Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mun
iform
distrib
ution
Set (|119868||119869|)
120572=085
120572=09
120572=095
Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
1(26)
27
897317
01
149
9906
01
107
9222
28
967373
01
155
9933
01
107
9167
27
106
7585
01
185
9987
01
107
9128
2(412
)47
169
7423
00
291
9919
00
207
9122
58
181
7438
01
303
9997
01
207
9363
50
199
7702
00
363
9982
00
207
9338
3(618
)72
258
7506
01
438
9976
01
312
9106
75279
8082
01
456
9996
01
312
9369
78304
8337
01
546
9997
01
312
9357
4(824)
99314
7956
01
553
9993
01
385
9206
111
338
8202
01
577
9999
01
385
9602
112
366
8492
01
697
9994
01
385
9620
5(1030)
132
415
8406
01
722
9896
01
512
9406
143
441
8322
02
752
9983
01
512
9692
142
476
8609
01
902
9993
01
512
9692
6(1236)
170
475
8249
02
843
9870
01
591
9367
181
504
8251
01
879
9982
01
591
9608
189
535
8726
01
1059
9997
01
591
9609
7(14
42)
215
573
8111
02
1006
9849
02
712
9606
243
609
8217
02
1048
9953
02
712
9589
233
644
8608
02
1258
9992
02
712
9582
8(1648)
248
686
7532
02
1185
9896
02
849
9725
284
727
8307
02
1233
9949
02
849
9613
281
768
9063
02
1473
9999
03
849
9622
9(1854)
313
691
8353
03
1236
9974
04
858
9709
351
731
8397
03
1290
9988
03
858
9765
352
760
8906
03
1560
9994
03
858
9758
10(2060)
372
794
8174
03
1409
9887
03
989
9657
428
843
8489
03
1469
9947
03
989
9773
432
892
9004
03
1769
9997
03
989
9738
11(2266
)451
830
8072
04
1493
9949
03
1031
9758
518
880
8328
04
1559
9982
03
1031
9723
516
930
8967
05
1889
9995
03
1031
9727
12(2472)
543
1025
7970
04
1771
9927
04
1267
9835
601
1083
8482
05
1843
9949
04
1267
9776
637
1146
9014
04
2203
9993
04
1267
9777
13(2678)
631
1026
8142
05
1819
9898
04
1274
9782
743
1087
8678
05
1897
9936
04
1274
9778
765
1146
9023
04
2287
9997
05
1273
9837
14(2884)
782
1101
7903
05
1949
9869
08
1361
9804
878
1165
8790
05
2033
9935
07
1361
9889
870
1227
9056
05
2453
9998
06
1361
9808
15(3090)
907
1091
8098
06
1997
9915
06
1367
9629
1018
1155
8329
06
2087
9981
06
1367
9807
1020
1225
8986
06
2537
9999
08
1367
9817
16(3296)
1062
1160
8798
07
2123
9944
07
1449
9787
1183
1229
8383
07
2219
9988
06
1449
9849
1193
1300
9105
06
2698
9998
06
1449
9815
17(34102)
1202
1298
7572
07
2328
9969
07
1614
9821
1377
1374
8860
08
2430
9997
07
1614
9824
1260
1452
9085
07
2940
9999
08
1614
9818
18(36108)
1394
1408
8121
08
2509
9947
08
1752
9809
1544
1492
8928
08
2617
9983
08
1752
9799
1548
1580
9173
08
3156
9999
08
1752
9809
19(38114
)1618
1508
8170
09
2676
9899
111877
9742
1760
1598
9130
09
2790
9934
08
1877
9908
1755
1688
9186
09
3360
9999
09
1877
9872
20(40120)
2206
1628
7719
102869
9989
102029
9873
2007
1732
8753
112990
9997
09
2029
9879
1904
1827
9138
103590
9998
09
2029
9875
21(42126)
2780
1622
8354
102900
9978
102017
9732
2609
1718
8854
103026
9876
102017
9818
2670
1812
9089
123655
9999
102017
9848
22(44132)
3189
1735
8494
113088
9985
152165
9857
2970
1840
8922
123220
9925
122165
9886
2873
1952
9206
113880
9999
112165
9850
23(46138)
3630
1937
8141
123372
9893
132406
9796
3222
2048
8937
123510
9934
142406
9906
3461
2170
8989
124199
10000
122406
9809
24(48144)
4226
1968
8226
133461
9932
132453
9814
3674
2085
8846
133605
9956
152453
9912
3556
2211
9300
134326
9999
132453
9819
25(50150)
4727
1919
8111
143456
9999
152407
9629
4019
2037
8699
143606
9999
142407
9857
4257
2161
9259
154356
9995
142407
9819
26(52156)
5300
1906
8096
153484
9878
172392
9658
4495
2021
8952
143640
9999
152392
9568
4486
2142
9182
1544
209999
152392
9854
27(54162)
5913
2096
8387
163751
9846
172617
9761
5117
2221
9045
163913
9998
172617
9544
4830
2355
9346
184723
9999
162617
9893
28(56168)
6310
2110
7758
173819
9933
192643
9829
5777
2234
9157
183987
9998
182643
9526
5550
2365
9045
184827
9999
172643
9831
29(58174)
6780
2175
7998
183937
9898
182720
9714
6303
2307
9022
20
4111
9969
192720
9623
6189
2444
9128
194981
10000
192720
9914
30(60180)
7593
2384
8321
194240
9967
20
2979
9841
6791
2527
8978
1944
199979
23
2979
9721
6878
2679
9143
21
5316
9999
20
2979
9869
31(62186)
8220
2281
7604
20
4157
9984
21
2856
9796
7734
2416
8934
21
4343
9997
21
2856
9679
7466
2557
9158
22
5273
9999
23
2856
9902
32(64192)
8766
2517
8378
21
4489
9932
25
3145
9729
8276
2671
8296
23
4680
9977
24
3145
9825
7897
2822
9089
24
5640
9999
22
3145
9891
33(66198)
10123
2532
7806
23
4555
9967
23
3169
9615
9010
2682
8947
24
4752
9989
24
3169
9703
9051
2835
9157
24
5742
9998
24
3169
9867
34(68204)
11010
2535
7934
24
4602
9999
29
3174
9631
9898
2685
8858
30
4806
9998
28
3174
9762
9863
2844
9236
26
5824
9999
25
3174
9889
35(70210)
11666
2714
8062
25
4862
9986
29
3391
9787
1066
62875
9137
26
5072
9999
29
3391
9716
10822
3043
9315
26
6122
9999
29
3391
9923
36(72216)
12011
2827
8104
26
5047
9991
28
3535
9798
1246
52984
9016
28
5263
9999
28
3535
9878
11667
3169
9253
30
6343
9999
29
3535
9896
37(74222)
11437
2826
8228
28
5096
9933
29
3543
9884
13582
3098
8895
29
5318
9999
32
3543
9888
12654
3178
9139
29
6427
10000
31
3543
9893
38(76228)
12435
2839
8152
29
5161
9897
29
3565
9788
13976
3103
8926
29
5388
9999
31
3565
9724
13837
3185
9255
33
6528
10000
31
3565
9895
39(78234)
1464
32945
8412
31
5331
9998
32
3695
9876
14794
3210
9005
31
5565
9999
34
3695
9713
14561
3313
9081
34
6735
10000
33
3695
9878
40(80240)
1604
22975
8510
32
5414
9957
36
3734
9796
16263
3346
8999
35
5654
9999
35
3734
9806
15869
3343
9248
35
6854
10000
35
3734
9873
Average
4482
1584
680
67
1228
397
9936
131978
796
82
4379
16913
8654
1329
626
9975
131978
79721
4295
1778
889
85
1335774
9997
131978
79775
Journal of Advanced Transportation 9
Table5Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mPo
isson
distr
ibution
set ( |119868 |
|119869 |)120572=
085120572=
09120572=
095Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
1(26)
27
907294
9967
9001
27
987770
9979
9013
24
103
9273
9999
9468
2(412
)46
170
7961
9921
9117
51
183
8305
9946
9128
49
201
9392
9998
9546
3(618
)68
258
7578
9938
8911
74280
8028
9968
9004
69
307
9182
9999
9658
4(824)
93315
8001
9928
9196
102
338
8438
9967
9172
92368
9468
9999
9613
5(1030)
120
413
8041
9939
9449
133
443
8447
9976
9272
121
479
9544
9999
9752
6(1236)
155
476
8084
9877
9554
167
508
8863
9991
9569
159
536
9456
9999
9734
7(14
42)
195
573
7876
9892
9663
213
609
9124
9998
9729
205
643
9314
10000
9768
8(1648)
232
687
7964
9997
9788
267
724
8523
9988
9687
238
772
9422
9998
9689
9(1854)
293
692
8267
9893
9809
327
730
8386
9999
9750
298
757
9228
9999
9755
10(2060)
356
796
7946
9899
9801
399
840
8299
9976
9686
360
885
9668
9999
9705
11(2266)
423
829
8201
9895
9756
489
877
8378
9987
9788
448
927
9476
9999
9812
12(2472)
512
1023
8338
9913
9853
554
1085
8501
9999
9881
514
1148
9398
9999
9888
13(2678)
592
1045
8416
9907
9873
687
1088
7998
9996
9802
607
1152
9504
9999
9821
14(2884)
723
1099
7946
9915
9788
838
1166
8056
9992
9829
776
1230
9546
9999
9839
15(3090)
855
1089
7915
9964
9703
942
1156
8432
9984
9778
903
1225
9223
9998
9785
16(3296)
1009
1158
8024
9897
9769
1163
1231
8500
9996
9824
1045
1222
9199
9999
9835
17(34102)
1165
1299
8034
9919
9779
1285
1375
8495
9937
9808
1190
1457
9258
10000
9846
18(36108)
1254
1411
8176
9943
9885
1479
1492
8561
9968
9868
1345
1579
9286
9999
9875
19(38114
)1504
1506
7970
9923
9842
1651
1594
8167
9972
9846
1564
1694
9376
9999
9786
20(40120)
1706
1635
7819
9916
9783
1854
1734
8607
9968
9776
1778
1825
9285
9904
9718
21(42126)
2054
1622
8354
9918
9834
2357
1720
8297
9978
9833
2226
1856
9501
9999
9834
22(44132)
2400
1736
8494
9922
9846
2852
1844
8269
9999
9816
2680
1946
9486
9999
9820
23(46138)
2651
1934
8141
9953
9793
3159
2049
8492
9969
9799
3072
2172
9358
10000
9813
24(48144)
3004
1961
8226
9947
9681
3522
2086
8513
9998
9706
3102
2209
9488
9999
9724
25(50150)
3308
1924
8111
9965
9746
3898
2038
8264
9999
9788
3639
2161
9546
9999
9813
26(52156)
3669
1911
8096
9939
9855
4305
2027
8722
9967
9874
4453
2142
9185
9999
9876
27(54162)
4418
2095
8387
9927
9754
4771
2219
8341
9989
9749
4806
2353
9289
9999
9755
28(56168)
5367
2113
8458
9968
9848
5356
2237
8452
9978
9855
5221
2369
9228
10000
9868
29(58174)
6217
2186
7998
9989
9694
5942
2315
8316
9996
9699
5853
2371
9167
9999
9705
30(60180)
7263
2382
8321
9991
9837
6527
2543
8359
9998
9886
6214
2678
9369
9997
9886
31(62186)
8214
2346
8404
9936
9888
7813
2430
8321
9999
9846
6876
2560
9358
9999
9847
32(64192)
8863
2518
8378
9938
9807
8106
2677
8356
9999
9833
7877
2822
9408
10000
9835
33(66198)
9985
2532
8406
9917
9709
9684
2689
8497
9996
9746
8088
2838
9507
9999
9749
34(68204)
11264
2534
8543
9908
9897
10986
2685
8809
9996
9782
8469
2843
8844
9998
9780
35(70210)
12035
2715
8258
9937
9799
12098
2886
8569
9991
9809
9946
3041
9459
9999
9813
36(72216)
12803
2825
8104
9929
9827
13076
3010
8651
9994
9834
11013
3170
9468
9999
9865
37(74222)
13021
2830
8228
9949
9826
13261
3009
8376
9999
9826
12014
3172
9346
9999
9888
38(76228)
13862
2843
8152
9933
9807
13995
3028
8760
9998
9841
12133
3187
9706
9999
9876
39(78234)
15387
2953
8312
9964
9799
14827
3133
9001
9995
9783
13381
3304
8913
9999
9799
40(80240)
16974
2975
8510
9985
9767
15256
3249
8217
9993
9781
15867
3347
9111
9999
9791
Average
4352
15875
8143
9934
9708
4362
16856
8437
9985
9707
3968
17763
9356
9997
9781
10 Journal of Advanced Transportation
Table6Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mno
rmaldistr
ibution
set ( |119868 |
|119869 |)120572=
085120572=
09120572=
095Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
1(26)
27
907294
9964
9117
27
947844
9970
9202
2558
106
8422
10000
9208
2(412
)46
170
7961
9921
9283
48
183
8328
9982
9283
4508
183
8823
10000
9370
3(618
)68
258
7578
9916
9228
68
280
7798
9971
9417
6803
307
7658
9998
9447
4(824)
93315
8000
9899
9367
96337
8379
9979
9467
9373
368
8438
9999
9579
5(1030)
120
413
7841
9927
9437
124
443
8763
9963
9476
1262
499
8850
9998
9708
6(1236)
155
476
7921
9919
9519
162
504
8124
9982
9582
16021
536
8954
10000
9756
7(14
42)
195
573
7976
9933
9598
196
609
8287
9975
9619
2011
646
8467
9999
9767
8(1648)
232
687
8031
9918
9705
235
727
8115
9965
9650
2404
772
8278
9999
9765
9(1854)
293
692
8086
9898
9759
291
734
8464
9974
9755
29815
796
8617
9998
9835
10(2060)
356
796
7984
9937
9770
358
842
8223
9988
9765
37973
921
8846
10000
9838
11(2266
)421
829
8101
9940
9755
436
877
8211
9986
9786
4514
8959
8497
9999
9789
12(2472)
514
1023
8033
9951
9785
501
1085
8162
9983
9788
52073
1147
8321
9999
9824
13(2678)
592
1023
8138
9916
9885
599
1087
8411
9970
9845
62937
1194
8609
9998
9795
14(2884)
728
1099
8235
9913
9915
722
1168
8435
9987
9806
75053
1270
8567
9997
9834
15(3090)
855
1089
8187
9922
9861
880
1157
8376
9976
9796
89222
1300
8611
9996
9798
16(3296)
1009
1158
8770
9897
9788
966
1231
8845
9984
9823
106288
1359
8997
9999
9848
17(34102)
1154
1299
8434
9925
9849
1137
1377
8601
9980
9798
123869
1493
8798
9996
9877
18(36108)
1407
1411
8321
9927
9866
1339
1495
8301
9965
9822
133807
1601
8567
9999
9834
19(38114
)1702
1506
8206
9915
9805
1672
1598
8977
9978
9845
152999
1694
8998
9999
9847
20(40120)
2091
1635
8528
9918
9889
1916
1737
8805
9976
9837
1890
021833
9012
9999
9791
21(42126)
2843
1622
8550
9934
9913
2432
1718
8886
9974
9813
2399
131815
9008
10000
9815
22(44132)
3300
1736
8449
9940
9859
2795
1843
8786
9977
9756
253106
1948
8997
9999
9826
23(46138)
3741
1934
8133
9911
9876
3143
2050
8876
9981
9785
274583
2173
9014
9999
9858
24(48144)
4531
1961
8245
9909
9846
3515
2090
8344
9993
9819
308089
2211
8661
9999
9883
25(50150)
4715
1924
8111
9929
9888
3959
2041
8279
9996
9837
3510
932158
8527
9999
9827
26(52156)
5223
1911
8096
9898
9857
4301
2025
8251
9997
9757
421747
2146
8497
9999
9914
27(54162)
5994
2095
8387
9908
9775
4835
2226
8497
9974
9780
4614
432359
8695
9999
9837
28(56168)
6406
2111
8411
9906
9798
5336
2240
8558
9991
9820
494113
2371
9249
9999
9816
29(58174)
6897
2176
8012
9916
9791
5948
2311
8211
9984
9765
624835
2445
8531
10000
9851
30(60180)
7436
2391
8111
9913
9766
6558
2533
8305
9968
9795
696116
2682
8575
9999
9799
31(62186)
8118
2282
8119
9928
9806
7069
2422
8417
9973
9873
797768
2564
8667
9999
9846
32(64192)
8895
2519
8223
9919
9757
7857
2675
8631
9982
9769
864813
2827
8976
10000
9905
33(66198)
11235
2533
8563
9928
9811
9162
2688
8547
9979
9828
973089
2843
8657
9999
9846
34(68204)
11587
2534
8249
9926
9847
9549
2686
8798
9938
9856
10523
3041
8999
10000
9877
35(70210)
1164
92715
8098
9895
9872
10345
2880
8602
9998
9869
11272
3052
9021
9999
9897
36(72216)
12224
2825
8403
9958
9770
11165
3002
8657
9999
9839
12456
3175
9113
9999
9864
37(74222)
12978
2830
8357
9937
9795
11850
3005
8496
9998
9894
13458
3176
8827
9999
9905
38(76228)
13426
2844
8267
9966
9759
12795
3019
8604
9999
9797
15155
3121
8978
9999
9853
39(78234)
15003
2955
8303
9956
9867
13816
3132
8588
9996
9885
15041
3310
9016
9999
9912
40(80240)
16495
2978
8257
9954
9826
14896
3160
8495
9997
9837
16563
3349
8957
9999
9933
Average
4619
15855
8174
9925
9742
4077
16828
8457
9981
9743
4350
17938
8732
9999
9794
Journal of Advanced Transportation 11
high service level requirement we recommend MIP-DF1 forsolutions
6 Conclusion
This paper investigates an ambulance location problem withambiguity set of demand in which the probability distribu-tion of the uncertain demand is unknown In such a data-driven environment only the mean and covariance demandsat emergency points are knownThe problem is to determinethe base locations and the number of ambulances at eachbase aiming at minimizing the total cost associated withlocating bases and assigning ambulances We propose adistribution-free model with chance constraints Then twoapproximated MIP formulations with different approxima-tion methods of chance constraints are proposed which arebased on different ambiguity sets of the unknown probabilitydistribution Numerical experiments on benchmarks and120 randomly generated instances are conducted and thecomputational results show that the first approximated MIPformulation is much conservative with overhigh system costwhile the second approximatedMIP formulation ismore costsaving with service level appropriately ensured
In future research wemay consider the following relevantissues First the uncertainty of driving time from a base toan emergency point shall be analysed Besides unexpected orsudden disasters may be taken into consideration Moreoverother uncertainties should be considered such as travel timethe number of available vehicles and so on The stochasticdependence should also further considered
Data Availability
The data of the benchmark instances in our manuscript canbe obtained in Nickel et al [8] For the test instances thedata is randomly generated and the way for generating data isstated in our manuscript The datasets generated during thecurrent study are available from the corresponding author onreasonable request
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This work was supported by the National Natural Sci-ence Foundation of China (NSFC) under Grants 7142800271531011 71571134 and 71771048 This work was also sup-ported by the Fundamental Research Funds for the CentralUniversities
References
[1] L Brotcorne G Laporte and F Semet ldquoAmbulance loca-tion and relocation modelsrdquo European Journal of OperationalResearch vol 147 no 3 pp 451ndash463 2003
[2] X Li Z Zhao X Zhu and T Wyatt ldquoCovering modelsand optimization techniques for emergency response facilitylocation and planning a reviewrdquo Mathematical Methods ofOperations Research vol 74 no 3 pp 281ndash310 2011
[3] C Toregas R Swain C ReVelle and L Bergman ldquoThe locationof emergency service facilitiesrdquoOperations Research vol 19 no6 pp 1363ndash1373 1971
[4] R Church and C ReVelle ldquoThe maximal covering locationproblemrdquo Papers in Regional Science vol 32 no 1 pp 101ndash1181974
[5] M Gendreau G Laporte and F Semet ldquoSolving an ambulancelocation model by tabu searchrdquo Location Science vol 5 no 2pp 75ndash88 1997
[6] A Ingolfsson S Budge and E Erkut ldquoOptimal ambulancelocation with random delays and travel timesrdquo Health CareManagement Science vol 11 no 3 pp 262ndash274 2008
[7] M Gendreau G Laporte and F Semet ldquoThemaximal expectedcoverage relocation problem for emergency vehiclesrdquo Journal ofthe Operational Research Society vol 57 no 1 pp 22ndash28 2006
[8] S Nickel M Reuter-Oppermann and F Saldanha-da-GamaldquoAmbulance location under stochastic demand A samplingapproachrdquo Operations Research for Health Care vol 8 pp 24ndash32 2016
[9] P Beraldi M E Bruni and D Conforti ldquoDesigning robustemergency medical service via stochastic programmingrdquo Euro-pean Journal of Operational Research vol 158 no 1 pp 183ndash1932004
[10] N Noyan ldquoAlternate risk measures for emergency medicalservice system designrdquo Annals of Operations Research vol 181pp 559ndash589 2010
[11] M R Wagner ldquoStochastic 0-1 linear programming underlimited distributional informationrdquoOperations Research Lettersvol 36 no 2 pp 150ndash156 2008
[12] EDelage andYYe ldquoDistributionally robust optimization undermoment uncertainty with application to data-driven problemsrdquoOperations Research vol 58 no 3 pp 595ndash612 2010
[13] J T Essen J L Hurink S Nickel and M Reuter ldquoModels forambulance planning on the strategic and the tactical levelrdquo BetaResearch School for Operations Management and Logistics p 272013
[14] E Erkut A Ingolfsson and G Erdogan ldquoAmbulance locationfor maximum survivalrdquoNaval Research Logistics (NRL) vol 55no 1 pp 42ndash58 2008
[15] S C Chapman and J A White ldquoProbabilistic formulations ofemergency service facilities location problemsrdquo in Proceedingsof ORSATIMS Conference San Juan Puerto Rico 1974
[16] A Ben-Tal D Den Hertog A De Waegenaere B Melenbergand G Rennen ldquoRobust solutions of optimization problemsaffected by uncertain probabilitiesrdquo Management Science vol59 no 2 pp 341ndash357 2013
[17] MNg ldquoDistribution-free vessel deployment for liner shippingrdquoEuropean Journal of Operational Research vol 238 no 3 pp858ndash862 2014
[18] M Ng ldquoContainer vessel fleet deployment for liner shippingwith stochastic dependencies in shipping demandrdquo Transporta-tion Research Part B Methodological vol 74 pp 79ndash87 2015
[19] R Jiang and Y Guan ldquoData-driven chance constrained stochas-tic programrdquo Mathematical Programming vol 158 no 1-2 SerA pp 291ndash327 2016
[20] C-M Lee and S-L Hsu ldquoThe effect of advertising on thedistribution-free newsboy problemrdquo International Journal ofProduction Economics vol 129 no 1 pp 217ndash224 2011
[21] K Kwon and T Cheong ldquoA minimax distribution-free proce-dure for a newsvendor problem with free shippingrdquo EuropeanJournal of Operational Research vol 232 no 1 pp 234ndash2402014
12 Journal of Advanced Transportation
[22] Y Zhang S Shen and S A Erdogan ldquoDistributionally robustappointment scheduling with moment-based ambiguity setrdquoOperations Research Letters vol 45 no 2 pp 139ndash144 2017
[23] Y Zhang S Shen and S A Erdogan ldquoSolving 0-1semidefinite programs for distributionally robust allocationof surgery blocksrdquo Social Science Electronic Publishing 2017httpsssrncomabstract=2997524
[24] F Zheng J He F Chu and M Liu ldquoA new distribution-freemodel for disassembly line balancing problem with stochastictask processing timesrdquo International Journal of ProductionResearch pp 1ndash13 2018
[25] J Cheng E Delage and A Lisser ldquoDistributionally robuststochastic knapsack problemrdquo SIAM Journal on Optimizationvol 24 no 3 pp 1485ndash1506 2014
[26] M Liu D Yang and F Hao ldquoOptimization for the locationsof ambulances under two-stage life rescue in the emergencymedical service a case study in Shanghai ChinardquoMathematicalProblems in Engineering vol 2017 Article ID 1830480 14 pages2017
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
8 Journal of Advanced Transportation
Table4Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mun
iform
distrib
ution
Set (|119868||119869|)
120572=085
120572=09
120572=095
Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
CTObj
Sel()
1(26)
27
897317
01
149
9906
01
107
9222
28
967373
01
155
9933
01
107
9167
27
106
7585
01
185
9987
01
107
9128
2(412
)47
169
7423
00
291
9919
00
207
9122
58
181
7438
01
303
9997
01
207
9363
50
199
7702
00
363
9982
00
207
9338
3(618
)72
258
7506
01
438
9976
01
312
9106
75279
8082
01
456
9996
01
312
9369
78304
8337
01
546
9997
01
312
9357
4(824)
99314
7956
01
553
9993
01
385
9206
111
338
8202
01
577
9999
01
385
9602
112
366
8492
01
697
9994
01
385
9620
5(1030)
132
415
8406
01
722
9896
01
512
9406
143
441
8322
02
752
9983
01
512
9692
142
476
8609
01
902
9993
01
512
9692
6(1236)
170
475
8249
02
843
9870
01
591
9367
181
504
8251
01
879
9982
01
591
9608
189
535
8726
01
1059
9997
01
591
9609
7(14
42)
215
573
8111
02
1006
9849
02
712
9606
243
609
8217
02
1048
9953
02
712
9589
233
644
8608
02
1258
9992
02
712
9582
8(1648)
248
686
7532
02
1185
9896
02
849
9725
284
727
8307
02
1233
9949
02
849
9613
281
768
9063
02
1473
9999
03
849
9622
9(1854)
313
691
8353
03
1236
9974
04
858
9709
351
731
8397
03
1290
9988
03
858
9765
352
760
8906
03
1560
9994
03
858
9758
10(2060)
372
794
8174
03
1409
9887
03
989
9657
428
843
8489
03
1469
9947
03
989
9773
432
892
9004
03
1769
9997
03
989
9738
11(2266
)451
830
8072
04
1493
9949
03
1031
9758
518
880
8328
04
1559
9982
03
1031
9723
516
930
8967
05
1889
9995
03
1031
9727
12(2472)
543
1025
7970
04
1771
9927
04
1267
9835
601
1083
8482
05
1843
9949
04
1267
9776
637
1146
9014
04
2203
9993
04
1267
9777
13(2678)
631
1026
8142
05
1819
9898
04
1274
9782
743
1087
8678
05
1897
9936
04
1274
9778
765
1146
9023
04
2287
9997
05
1273
9837
14(2884)
782
1101
7903
05
1949
9869
08
1361
9804
878
1165
8790
05
2033
9935
07
1361
9889
870
1227
9056
05
2453
9998
06
1361
9808
15(3090)
907
1091
8098
06
1997
9915
06
1367
9629
1018
1155
8329
06
2087
9981
06
1367
9807
1020
1225
8986
06
2537
9999
08
1367
9817
16(3296)
1062
1160
8798
07
2123
9944
07
1449
9787
1183
1229
8383
07
2219
9988
06
1449
9849
1193
1300
9105
06
2698
9998
06
1449
9815
17(34102)
1202
1298
7572
07
2328
9969
07
1614
9821
1377
1374
8860
08
2430
9997
07
1614
9824
1260
1452
9085
07
2940
9999
08
1614
9818
18(36108)
1394
1408
8121
08
2509
9947
08
1752
9809
1544
1492
8928
08
2617
9983
08
1752
9799
1548
1580
9173
08
3156
9999
08
1752
9809
19(38114
)1618
1508
8170
09
2676
9899
111877
9742
1760
1598
9130
09
2790
9934
08
1877
9908
1755
1688
9186
09
3360
9999
09
1877
9872
20(40120)
2206
1628
7719
102869
9989
102029
9873
2007
1732
8753
112990
9997
09
2029
9879
1904
1827
9138
103590
9998
09
2029
9875
21(42126)
2780
1622
8354
102900
9978
102017
9732
2609
1718
8854
103026
9876
102017
9818
2670
1812
9089
123655
9999
102017
9848
22(44132)
3189
1735
8494
113088
9985
152165
9857
2970
1840
8922
123220
9925
122165
9886
2873
1952
9206
113880
9999
112165
9850
23(46138)
3630
1937
8141
123372
9893
132406
9796
3222
2048
8937
123510
9934
142406
9906
3461
2170
8989
124199
10000
122406
9809
24(48144)
4226
1968
8226
133461
9932
132453
9814
3674
2085
8846
133605
9956
152453
9912
3556
2211
9300
134326
9999
132453
9819
25(50150)
4727
1919
8111
143456
9999
152407
9629
4019
2037
8699
143606
9999
142407
9857
4257
2161
9259
154356
9995
142407
9819
26(52156)
5300
1906
8096
153484
9878
172392
9658
4495
2021
8952
143640
9999
152392
9568
4486
2142
9182
1544
209999
152392
9854
27(54162)
5913
2096
8387
163751
9846
172617
9761
5117
2221
9045
163913
9998
172617
9544
4830
2355
9346
184723
9999
162617
9893
28(56168)
6310
2110
7758
173819
9933
192643
9829
5777
2234
9157
183987
9998
182643
9526
5550
2365
9045
184827
9999
172643
9831
29(58174)
6780
2175
7998
183937
9898
182720
9714
6303
2307
9022
20
4111
9969
192720
9623
6189
2444
9128
194981
10000
192720
9914
30(60180)
7593
2384
8321
194240
9967
20
2979
9841
6791
2527
8978
1944
199979
23
2979
9721
6878
2679
9143
21
5316
9999
20
2979
9869
31(62186)
8220
2281
7604
20
4157
9984
21
2856
9796
7734
2416
8934
21
4343
9997
21
2856
9679
7466
2557
9158
22
5273
9999
23
2856
9902
32(64192)
8766
2517
8378
21
4489
9932
25
3145
9729
8276
2671
8296
23
4680
9977
24
3145
9825
7897
2822
9089
24
5640
9999
22
3145
9891
33(66198)
10123
2532
7806
23
4555
9967
23
3169
9615
9010
2682
8947
24
4752
9989
24
3169
9703
9051
2835
9157
24
5742
9998
24
3169
9867
34(68204)
11010
2535
7934
24
4602
9999
29
3174
9631
9898
2685
8858
30
4806
9998
28
3174
9762
9863
2844
9236
26
5824
9999
25
3174
9889
35(70210)
11666
2714
8062
25
4862
9986
29
3391
9787
1066
62875
9137
26
5072
9999
29
3391
9716
10822
3043
9315
26
6122
9999
29
3391
9923
36(72216)
12011
2827
8104
26
5047
9991
28
3535
9798
1246
52984
9016
28
5263
9999
28
3535
9878
11667
3169
9253
30
6343
9999
29
3535
9896
37(74222)
11437
2826
8228
28
5096
9933
29
3543
9884
13582
3098
8895
29
5318
9999
32
3543
9888
12654
3178
9139
29
6427
10000
31
3543
9893
38(76228)
12435
2839
8152
29
5161
9897
29
3565
9788
13976
3103
8926
29
5388
9999
31
3565
9724
13837
3185
9255
33
6528
10000
31
3565
9895
39(78234)
1464
32945
8412
31
5331
9998
32
3695
9876
14794
3210
9005
31
5565
9999
34
3695
9713
14561
3313
9081
34
6735
10000
33
3695
9878
40(80240)
1604
22975
8510
32
5414
9957
36
3734
9796
16263
3346
8999
35
5654
9999
35
3734
9806
15869
3343
9248
35
6854
10000
35
3734
9873
Average
4482
1584
680
67
1228
397
9936
131978
796
82
4379
16913
8654
1329
626
9975
131978
79721
4295
1778
889
85
1335774
9997
131978
79775
Journal of Advanced Transportation 9
Table5Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mPo
isson
distr
ibution
set ( |119868 |
|119869 |)120572=
085120572=
09120572=
095Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
1(26)
27
907294
9967
9001
27
987770
9979
9013
24
103
9273
9999
9468
2(412
)46
170
7961
9921
9117
51
183
8305
9946
9128
49
201
9392
9998
9546
3(618
)68
258
7578
9938
8911
74280
8028
9968
9004
69
307
9182
9999
9658
4(824)
93315
8001
9928
9196
102
338
8438
9967
9172
92368
9468
9999
9613
5(1030)
120
413
8041
9939
9449
133
443
8447
9976
9272
121
479
9544
9999
9752
6(1236)
155
476
8084
9877
9554
167
508
8863
9991
9569
159
536
9456
9999
9734
7(14
42)
195
573
7876
9892
9663
213
609
9124
9998
9729
205
643
9314
10000
9768
8(1648)
232
687
7964
9997
9788
267
724
8523
9988
9687
238
772
9422
9998
9689
9(1854)
293
692
8267
9893
9809
327
730
8386
9999
9750
298
757
9228
9999
9755
10(2060)
356
796
7946
9899
9801
399
840
8299
9976
9686
360
885
9668
9999
9705
11(2266)
423
829
8201
9895
9756
489
877
8378
9987
9788
448
927
9476
9999
9812
12(2472)
512
1023
8338
9913
9853
554
1085
8501
9999
9881
514
1148
9398
9999
9888
13(2678)
592
1045
8416
9907
9873
687
1088
7998
9996
9802
607
1152
9504
9999
9821
14(2884)
723
1099
7946
9915
9788
838
1166
8056
9992
9829
776
1230
9546
9999
9839
15(3090)
855
1089
7915
9964
9703
942
1156
8432
9984
9778
903
1225
9223
9998
9785
16(3296)
1009
1158
8024
9897
9769
1163
1231
8500
9996
9824
1045
1222
9199
9999
9835
17(34102)
1165
1299
8034
9919
9779
1285
1375
8495
9937
9808
1190
1457
9258
10000
9846
18(36108)
1254
1411
8176
9943
9885
1479
1492
8561
9968
9868
1345
1579
9286
9999
9875
19(38114
)1504
1506
7970
9923
9842
1651
1594
8167
9972
9846
1564
1694
9376
9999
9786
20(40120)
1706
1635
7819
9916
9783
1854
1734
8607
9968
9776
1778
1825
9285
9904
9718
21(42126)
2054
1622
8354
9918
9834
2357
1720
8297
9978
9833
2226
1856
9501
9999
9834
22(44132)
2400
1736
8494
9922
9846
2852
1844
8269
9999
9816
2680
1946
9486
9999
9820
23(46138)
2651
1934
8141
9953
9793
3159
2049
8492
9969
9799
3072
2172
9358
10000
9813
24(48144)
3004
1961
8226
9947
9681
3522
2086
8513
9998
9706
3102
2209
9488
9999
9724
25(50150)
3308
1924
8111
9965
9746
3898
2038
8264
9999
9788
3639
2161
9546
9999
9813
26(52156)
3669
1911
8096
9939
9855
4305
2027
8722
9967
9874
4453
2142
9185
9999
9876
27(54162)
4418
2095
8387
9927
9754
4771
2219
8341
9989
9749
4806
2353
9289
9999
9755
28(56168)
5367
2113
8458
9968
9848
5356
2237
8452
9978
9855
5221
2369
9228
10000
9868
29(58174)
6217
2186
7998
9989
9694
5942
2315
8316
9996
9699
5853
2371
9167
9999
9705
30(60180)
7263
2382
8321
9991
9837
6527
2543
8359
9998
9886
6214
2678
9369
9997
9886
31(62186)
8214
2346
8404
9936
9888
7813
2430
8321
9999
9846
6876
2560
9358
9999
9847
32(64192)
8863
2518
8378
9938
9807
8106
2677
8356
9999
9833
7877
2822
9408
10000
9835
33(66198)
9985
2532
8406
9917
9709
9684
2689
8497
9996
9746
8088
2838
9507
9999
9749
34(68204)
11264
2534
8543
9908
9897
10986
2685
8809
9996
9782
8469
2843
8844
9998
9780
35(70210)
12035
2715
8258
9937
9799
12098
2886
8569
9991
9809
9946
3041
9459
9999
9813
36(72216)
12803
2825
8104
9929
9827
13076
3010
8651
9994
9834
11013
3170
9468
9999
9865
37(74222)
13021
2830
8228
9949
9826
13261
3009
8376
9999
9826
12014
3172
9346
9999
9888
38(76228)
13862
2843
8152
9933
9807
13995
3028
8760
9998
9841
12133
3187
9706
9999
9876
39(78234)
15387
2953
8312
9964
9799
14827
3133
9001
9995
9783
13381
3304
8913
9999
9799
40(80240)
16974
2975
8510
9985
9767
15256
3249
8217
9993
9781
15867
3347
9111
9999
9791
Average
4352
15875
8143
9934
9708
4362
16856
8437
9985
9707
3968
17763
9356
9997
9781
10 Journal of Advanced Transportation
Table6Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mno
rmaldistr
ibution
set ( |119868 |
|119869 |)120572=
085120572=
09120572=
095Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
1(26)
27
907294
9964
9117
27
947844
9970
9202
2558
106
8422
10000
9208
2(412
)46
170
7961
9921
9283
48
183
8328
9982
9283
4508
183
8823
10000
9370
3(618
)68
258
7578
9916
9228
68
280
7798
9971
9417
6803
307
7658
9998
9447
4(824)
93315
8000
9899
9367
96337
8379
9979
9467
9373
368
8438
9999
9579
5(1030)
120
413
7841
9927
9437
124
443
8763
9963
9476
1262
499
8850
9998
9708
6(1236)
155
476
7921
9919
9519
162
504
8124
9982
9582
16021
536
8954
10000
9756
7(14
42)
195
573
7976
9933
9598
196
609
8287
9975
9619
2011
646
8467
9999
9767
8(1648)
232
687
8031
9918
9705
235
727
8115
9965
9650
2404
772
8278
9999
9765
9(1854)
293
692
8086
9898
9759
291
734
8464
9974
9755
29815
796
8617
9998
9835
10(2060)
356
796
7984
9937
9770
358
842
8223
9988
9765
37973
921
8846
10000
9838
11(2266
)421
829
8101
9940
9755
436
877
8211
9986
9786
4514
8959
8497
9999
9789
12(2472)
514
1023
8033
9951
9785
501
1085
8162
9983
9788
52073
1147
8321
9999
9824
13(2678)
592
1023
8138
9916
9885
599
1087
8411
9970
9845
62937
1194
8609
9998
9795
14(2884)
728
1099
8235
9913
9915
722
1168
8435
9987
9806
75053
1270
8567
9997
9834
15(3090)
855
1089
8187
9922
9861
880
1157
8376
9976
9796
89222
1300
8611
9996
9798
16(3296)
1009
1158
8770
9897
9788
966
1231
8845
9984
9823
106288
1359
8997
9999
9848
17(34102)
1154
1299
8434
9925
9849
1137
1377
8601
9980
9798
123869
1493
8798
9996
9877
18(36108)
1407
1411
8321
9927
9866
1339
1495
8301
9965
9822
133807
1601
8567
9999
9834
19(38114
)1702
1506
8206
9915
9805
1672
1598
8977
9978
9845
152999
1694
8998
9999
9847
20(40120)
2091
1635
8528
9918
9889
1916
1737
8805
9976
9837
1890
021833
9012
9999
9791
21(42126)
2843
1622
8550
9934
9913
2432
1718
8886
9974
9813
2399
131815
9008
10000
9815
22(44132)
3300
1736
8449
9940
9859
2795
1843
8786
9977
9756
253106
1948
8997
9999
9826
23(46138)
3741
1934
8133
9911
9876
3143
2050
8876
9981
9785
274583
2173
9014
9999
9858
24(48144)
4531
1961
8245
9909
9846
3515
2090
8344
9993
9819
308089
2211
8661
9999
9883
25(50150)
4715
1924
8111
9929
9888
3959
2041
8279
9996
9837
3510
932158
8527
9999
9827
26(52156)
5223
1911
8096
9898
9857
4301
2025
8251
9997
9757
421747
2146
8497
9999
9914
27(54162)
5994
2095
8387
9908
9775
4835
2226
8497
9974
9780
4614
432359
8695
9999
9837
28(56168)
6406
2111
8411
9906
9798
5336
2240
8558
9991
9820
494113
2371
9249
9999
9816
29(58174)
6897
2176
8012
9916
9791
5948
2311
8211
9984
9765
624835
2445
8531
10000
9851
30(60180)
7436
2391
8111
9913
9766
6558
2533
8305
9968
9795
696116
2682
8575
9999
9799
31(62186)
8118
2282
8119
9928
9806
7069
2422
8417
9973
9873
797768
2564
8667
9999
9846
32(64192)
8895
2519
8223
9919
9757
7857
2675
8631
9982
9769
864813
2827
8976
10000
9905
33(66198)
11235
2533
8563
9928
9811
9162
2688
8547
9979
9828
973089
2843
8657
9999
9846
34(68204)
11587
2534
8249
9926
9847
9549
2686
8798
9938
9856
10523
3041
8999
10000
9877
35(70210)
1164
92715
8098
9895
9872
10345
2880
8602
9998
9869
11272
3052
9021
9999
9897
36(72216)
12224
2825
8403
9958
9770
11165
3002
8657
9999
9839
12456
3175
9113
9999
9864
37(74222)
12978
2830
8357
9937
9795
11850
3005
8496
9998
9894
13458
3176
8827
9999
9905
38(76228)
13426
2844
8267
9966
9759
12795
3019
8604
9999
9797
15155
3121
8978
9999
9853
39(78234)
15003
2955
8303
9956
9867
13816
3132
8588
9996
9885
15041
3310
9016
9999
9912
40(80240)
16495
2978
8257
9954
9826
14896
3160
8495
9997
9837
16563
3349
8957
9999
9933
Average
4619
15855
8174
9925
9742
4077
16828
8457
9981
9743
4350
17938
8732
9999
9794
Journal of Advanced Transportation 11
high service level requirement we recommend MIP-DF1 forsolutions
6 Conclusion
This paper investigates an ambulance location problem withambiguity set of demand in which the probability distribu-tion of the uncertain demand is unknown In such a data-driven environment only the mean and covariance demandsat emergency points are knownThe problem is to determinethe base locations and the number of ambulances at eachbase aiming at minimizing the total cost associated withlocating bases and assigning ambulances We propose adistribution-free model with chance constraints Then twoapproximated MIP formulations with different approxima-tion methods of chance constraints are proposed which arebased on different ambiguity sets of the unknown probabilitydistribution Numerical experiments on benchmarks and120 randomly generated instances are conducted and thecomputational results show that the first approximated MIPformulation is much conservative with overhigh system costwhile the second approximatedMIP formulation ismore costsaving with service level appropriately ensured
In future research wemay consider the following relevantissues First the uncertainty of driving time from a base toan emergency point shall be analysed Besides unexpected orsudden disasters may be taken into consideration Moreoverother uncertainties should be considered such as travel timethe number of available vehicles and so on The stochasticdependence should also further considered
Data Availability
The data of the benchmark instances in our manuscript canbe obtained in Nickel et al [8] For the test instances thedata is randomly generated and the way for generating data isstated in our manuscript The datasets generated during thecurrent study are available from the corresponding author onreasonable request
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This work was supported by the National Natural Sci-ence Foundation of China (NSFC) under Grants 7142800271531011 71571134 and 71771048 This work was also sup-ported by the Fundamental Research Funds for the CentralUniversities
References
[1] L Brotcorne G Laporte and F Semet ldquoAmbulance loca-tion and relocation modelsrdquo European Journal of OperationalResearch vol 147 no 3 pp 451ndash463 2003
[2] X Li Z Zhao X Zhu and T Wyatt ldquoCovering modelsand optimization techniques for emergency response facilitylocation and planning a reviewrdquo Mathematical Methods ofOperations Research vol 74 no 3 pp 281ndash310 2011
[3] C Toregas R Swain C ReVelle and L Bergman ldquoThe locationof emergency service facilitiesrdquoOperations Research vol 19 no6 pp 1363ndash1373 1971
[4] R Church and C ReVelle ldquoThe maximal covering locationproblemrdquo Papers in Regional Science vol 32 no 1 pp 101ndash1181974
[5] M Gendreau G Laporte and F Semet ldquoSolving an ambulancelocation model by tabu searchrdquo Location Science vol 5 no 2pp 75ndash88 1997
[6] A Ingolfsson S Budge and E Erkut ldquoOptimal ambulancelocation with random delays and travel timesrdquo Health CareManagement Science vol 11 no 3 pp 262ndash274 2008
[7] M Gendreau G Laporte and F Semet ldquoThemaximal expectedcoverage relocation problem for emergency vehiclesrdquo Journal ofthe Operational Research Society vol 57 no 1 pp 22ndash28 2006
[8] S Nickel M Reuter-Oppermann and F Saldanha-da-GamaldquoAmbulance location under stochastic demand A samplingapproachrdquo Operations Research for Health Care vol 8 pp 24ndash32 2016
[9] P Beraldi M E Bruni and D Conforti ldquoDesigning robustemergency medical service via stochastic programmingrdquo Euro-pean Journal of Operational Research vol 158 no 1 pp 183ndash1932004
[10] N Noyan ldquoAlternate risk measures for emergency medicalservice system designrdquo Annals of Operations Research vol 181pp 559ndash589 2010
[11] M R Wagner ldquoStochastic 0-1 linear programming underlimited distributional informationrdquoOperations Research Lettersvol 36 no 2 pp 150ndash156 2008
[12] EDelage andYYe ldquoDistributionally robust optimization undermoment uncertainty with application to data-driven problemsrdquoOperations Research vol 58 no 3 pp 595ndash612 2010
[13] J T Essen J L Hurink S Nickel and M Reuter ldquoModels forambulance planning on the strategic and the tactical levelrdquo BetaResearch School for Operations Management and Logistics p 272013
[14] E Erkut A Ingolfsson and G Erdogan ldquoAmbulance locationfor maximum survivalrdquoNaval Research Logistics (NRL) vol 55no 1 pp 42ndash58 2008
[15] S C Chapman and J A White ldquoProbabilistic formulations ofemergency service facilities location problemsrdquo in Proceedingsof ORSATIMS Conference San Juan Puerto Rico 1974
[16] A Ben-Tal D Den Hertog A De Waegenaere B Melenbergand G Rennen ldquoRobust solutions of optimization problemsaffected by uncertain probabilitiesrdquo Management Science vol59 no 2 pp 341ndash357 2013
[17] MNg ldquoDistribution-free vessel deployment for liner shippingrdquoEuropean Journal of Operational Research vol 238 no 3 pp858ndash862 2014
[18] M Ng ldquoContainer vessel fleet deployment for liner shippingwith stochastic dependencies in shipping demandrdquo Transporta-tion Research Part B Methodological vol 74 pp 79ndash87 2015
[19] R Jiang and Y Guan ldquoData-driven chance constrained stochas-tic programrdquo Mathematical Programming vol 158 no 1-2 SerA pp 291ndash327 2016
[20] C-M Lee and S-L Hsu ldquoThe effect of advertising on thedistribution-free newsboy problemrdquo International Journal ofProduction Economics vol 129 no 1 pp 217ndash224 2011
[21] K Kwon and T Cheong ldquoA minimax distribution-free proce-dure for a newsvendor problem with free shippingrdquo EuropeanJournal of Operational Research vol 232 no 1 pp 234ndash2402014
12 Journal of Advanced Transportation
[22] Y Zhang S Shen and S A Erdogan ldquoDistributionally robustappointment scheduling with moment-based ambiguity setrdquoOperations Research Letters vol 45 no 2 pp 139ndash144 2017
[23] Y Zhang S Shen and S A Erdogan ldquoSolving 0-1semidefinite programs for distributionally robust allocationof surgery blocksrdquo Social Science Electronic Publishing 2017httpsssrncomabstract=2997524
[24] F Zheng J He F Chu and M Liu ldquoA new distribution-freemodel for disassembly line balancing problem with stochastictask processing timesrdquo International Journal of ProductionResearch pp 1ndash13 2018
[25] J Cheng E Delage and A Lisser ldquoDistributionally robuststochastic knapsack problemrdquo SIAM Journal on Optimizationvol 24 no 3 pp 1485ndash1506 2014
[26] M Liu D Yang and F Hao ldquoOptimization for the locationsof ambulances under two-stage life rescue in the emergencymedical service a case study in Shanghai ChinardquoMathematicalProblems in Engineering vol 2017 Article ID 1830480 14 pages2017
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
Journal of Advanced Transportation 9
Table5Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mPo
isson
distr
ibution
set ( |119868 |
|119869 |)120572=
085120572=
09120572=
095Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
1(26)
27
907294
9967
9001
27
987770
9979
9013
24
103
9273
9999
9468
2(412
)46
170
7961
9921
9117
51
183
8305
9946
9128
49
201
9392
9998
9546
3(618
)68
258
7578
9938
8911
74280
8028
9968
9004
69
307
9182
9999
9658
4(824)
93315
8001
9928
9196
102
338
8438
9967
9172
92368
9468
9999
9613
5(1030)
120
413
8041
9939
9449
133
443
8447
9976
9272
121
479
9544
9999
9752
6(1236)
155
476
8084
9877
9554
167
508
8863
9991
9569
159
536
9456
9999
9734
7(14
42)
195
573
7876
9892
9663
213
609
9124
9998
9729
205
643
9314
10000
9768
8(1648)
232
687
7964
9997
9788
267
724
8523
9988
9687
238
772
9422
9998
9689
9(1854)
293
692
8267
9893
9809
327
730
8386
9999
9750
298
757
9228
9999
9755
10(2060)
356
796
7946
9899
9801
399
840
8299
9976
9686
360
885
9668
9999
9705
11(2266)
423
829
8201
9895
9756
489
877
8378
9987
9788
448
927
9476
9999
9812
12(2472)
512
1023
8338
9913
9853
554
1085
8501
9999
9881
514
1148
9398
9999
9888
13(2678)
592
1045
8416
9907
9873
687
1088
7998
9996
9802
607
1152
9504
9999
9821
14(2884)
723
1099
7946
9915
9788
838
1166
8056
9992
9829
776
1230
9546
9999
9839
15(3090)
855
1089
7915
9964
9703
942
1156
8432
9984
9778
903
1225
9223
9998
9785
16(3296)
1009
1158
8024
9897
9769
1163
1231
8500
9996
9824
1045
1222
9199
9999
9835
17(34102)
1165
1299
8034
9919
9779
1285
1375
8495
9937
9808
1190
1457
9258
10000
9846
18(36108)
1254
1411
8176
9943
9885
1479
1492
8561
9968
9868
1345
1579
9286
9999
9875
19(38114
)1504
1506
7970
9923
9842
1651
1594
8167
9972
9846
1564
1694
9376
9999
9786
20(40120)
1706
1635
7819
9916
9783
1854
1734
8607
9968
9776
1778
1825
9285
9904
9718
21(42126)
2054
1622
8354
9918
9834
2357
1720
8297
9978
9833
2226
1856
9501
9999
9834
22(44132)
2400
1736
8494
9922
9846
2852
1844
8269
9999
9816
2680
1946
9486
9999
9820
23(46138)
2651
1934
8141
9953
9793
3159
2049
8492
9969
9799
3072
2172
9358
10000
9813
24(48144)
3004
1961
8226
9947
9681
3522
2086
8513
9998
9706
3102
2209
9488
9999
9724
25(50150)
3308
1924
8111
9965
9746
3898
2038
8264
9999
9788
3639
2161
9546
9999
9813
26(52156)
3669
1911
8096
9939
9855
4305
2027
8722
9967
9874
4453
2142
9185
9999
9876
27(54162)
4418
2095
8387
9927
9754
4771
2219
8341
9989
9749
4806
2353
9289
9999
9755
28(56168)
5367
2113
8458
9968
9848
5356
2237
8452
9978
9855
5221
2369
9228
10000
9868
29(58174)
6217
2186
7998
9989
9694
5942
2315
8316
9996
9699
5853
2371
9167
9999
9705
30(60180)
7263
2382
8321
9991
9837
6527
2543
8359
9998
9886
6214
2678
9369
9997
9886
31(62186)
8214
2346
8404
9936
9888
7813
2430
8321
9999
9846
6876
2560
9358
9999
9847
32(64192)
8863
2518
8378
9938
9807
8106
2677
8356
9999
9833
7877
2822
9408
10000
9835
33(66198)
9985
2532
8406
9917
9709
9684
2689
8497
9996
9746
8088
2838
9507
9999
9749
34(68204)
11264
2534
8543
9908
9897
10986
2685
8809
9996
9782
8469
2843
8844
9998
9780
35(70210)
12035
2715
8258
9937
9799
12098
2886
8569
9991
9809
9946
3041
9459
9999
9813
36(72216)
12803
2825
8104
9929
9827
13076
3010
8651
9994
9834
11013
3170
9468
9999
9865
37(74222)
13021
2830
8228
9949
9826
13261
3009
8376
9999
9826
12014
3172
9346
9999
9888
38(76228)
13862
2843
8152
9933
9807
13995
3028
8760
9998
9841
12133
3187
9706
9999
9876
39(78234)
15387
2953
8312
9964
9799
14827
3133
9001
9995
9783
13381
3304
8913
9999
9799
40(80240)
16974
2975
8510
9985
9767
15256
3249
8217
9993
9781
15867
3347
9111
9999
9791
Average
4352
15875
8143
9934
9708
4362
16856
8437
9985
9707
3968
17763
9356
9997
9781
10 Journal of Advanced Transportation
Table6Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mno
rmaldistr
ibution
set ( |119868 |
|119869 |)120572=
085120572=
09120572=
095Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
1(26)
27
907294
9964
9117
27
947844
9970
9202
2558
106
8422
10000
9208
2(412
)46
170
7961
9921
9283
48
183
8328
9982
9283
4508
183
8823
10000
9370
3(618
)68
258
7578
9916
9228
68
280
7798
9971
9417
6803
307
7658
9998
9447
4(824)
93315
8000
9899
9367
96337
8379
9979
9467
9373
368
8438
9999
9579
5(1030)
120
413
7841
9927
9437
124
443
8763
9963
9476
1262
499
8850
9998
9708
6(1236)
155
476
7921
9919
9519
162
504
8124
9982
9582
16021
536
8954
10000
9756
7(14
42)
195
573
7976
9933
9598
196
609
8287
9975
9619
2011
646
8467
9999
9767
8(1648)
232
687
8031
9918
9705
235
727
8115
9965
9650
2404
772
8278
9999
9765
9(1854)
293
692
8086
9898
9759
291
734
8464
9974
9755
29815
796
8617
9998
9835
10(2060)
356
796
7984
9937
9770
358
842
8223
9988
9765
37973
921
8846
10000
9838
11(2266
)421
829
8101
9940
9755
436
877
8211
9986
9786
4514
8959
8497
9999
9789
12(2472)
514
1023
8033
9951
9785
501
1085
8162
9983
9788
52073
1147
8321
9999
9824
13(2678)
592
1023
8138
9916
9885
599
1087
8411
9970
9845
62937
1194
8609
9998
9795
14(2884)
728
1099
8235
9913
9915
722
1168
8435
9987
9806
75053
1270
8567
9997
9834
15(3090)
855
1089
8187
9922
9861
880
1157
8376
9976
9796
89222
1300
8611
9996
9798
16(3296)
1009
1158
8770
9897
9788
966
1231
8845
9984
9823
106288
1359
8997
9999
9848
17(34102)
1154
1299
8434
9925
9849
1137
1377
8601
9980
9798
123869
1493
8798
9996
9877
18(36108)
1407
1411
8321
9927
9866
1339
1495
8301
9965
9822
133807
1601
8567
9999
9834
19(38114
)1702
1506
8206
9915
9805
1672
1598
8977
9978
9845
152999
1694
8998
9999
9847
20(40120)
2091
1635
8528
9918
9889
1916
1737
8805
9976
9837
1890
021833
9012
9999
9791
21(42126)
2843
1622
8550
9934
9913
2432
1718
8886
9974
9813
2399
131815
9008
10000
9815
22(44132)
3300
1736
8449
9940
9859
2795
1843
8786
9977
9756
253106
1948
8997
9999
9826
23(46138)
3741
1934
8133
9911
9876
3143
2050
8876
9981
9785
274583
2173
9014
9999
9858
24(48144)
4531
1961
8245
9909
9846
3515
2090
8344
9993
9819
308089
2211
8661
9999
9883
25(50150)
4715
1924
8111
9929
9888
3959
2041
8279
9996
9837
3510
932158
8527
9999
9827
26(52156)
5223
1911
8096
9898
9857
4301
2025
8251
9997
9757
421747
2146
8497
9999
9914
27(54162)
5994
2095
8387
9908
9775
4835
2226
8497
9974
9780
4614
432359
8695
9999
9837
28(56168)
6406
2111
8411
9906
9798
5336
2240
8558
9991
9820
494113
2371
9249
9999
9816
29(58174)
6897
2176
8012
9916
9791
5948
2311
8211
9984
9765
624835
2445
8531
10000
9851
30(60180)
7436
2391
8111
9913
9766
6558
2533
8305
9968
9795
696116
2682
8575
9999
9799
31(62186)
8118
2282
8119
9928
9806
7069
2422
8417
9973
9873
797768
2564
8667
9999
9846
32(64192)
8895
2519
8223
9919
9757
7857
2675
8631
9982
9769
864813
2827
8976
10000
9905
33(66198)
11235
2533
8563
9928
9811
9162
2688
8547
9979
9828
973089
2843
8657
9999
9846
34(68204)
11587
2534
8249
9926
9847
9549
2686
8798
9938
9856
10523
3041
8999
10000
9877
35(70210)
1164
92715
8098
9895
9872
10345
2880
8602
9998
9869
11272
3052
9021
9999
9897
36(72216)
12224
2825
8403
9958
9770
11165
3002
8657
9999
9839
12456
3175
9113
9999
9864
37(74222)
12978
2830
8357
9937
9795
11850
3005
8496
9998
9894
13458
3176
8827
9999
9905
38(76228)
13426
2844
8267
9966
9759
12795
3019
8604
9999
9797
15155
3121
8978
9999
9853
39(78234)
15003
2955
8303
9956
9867
13816
3132
8588
9996
9885
15041
3310
9016
9999
9912
40(80240)
16495
2978
8257
9954
9826
14896
3160
8495
9997
9837
16563
3349
8957
9999
9933
Average
4619
15855
8174
9925
9742
4077
16828
8457
9981
9743
4350
17938
8732
9999
9794
Journal of Advanced Transportation 11
high service level requirement we recommend MIP-DF1 forsolutions
6 Conclusion
This paper investigates an ambulance location problem withambiguity set of demand in which the probability distribu-tion of the uncertain demand is unknown In such a data-driven environment only the mean and covariance demandsat emergency points are knownThe problem is to determinethe base locations and the number of ambulances at eachbase aiming at minimizing the total cost associated withlocating bases and assigning ambulances We propose adistribution-free model with chance constraints Then twoapproximated MIP formulations with different approxima-tion methods of chance constraints are proposed which arebased on different ambiguity sets of the unknown probabilitydistribution Numerical experiments on benchmarks and120 randomly generated instances are conducted and thecomputational results show that the first approximated MIPformulation is much conservative with overhigh system costwhile the second approximatedMIP formulation ismore costsaving with service level appropriately ensured
In future research wemay consider the following relevantissues First the uncertainty of driving time from a base toan emergency point shall be analysed Besides unexpected orsudden disasters may be taken into consideration Moreoverother uncertainties should be considered such as travel timethe number of available vehicles and so on The stochasticdependence should also further considered
Data Availability
The data of the benchmark instances in our manuscript canbe obtained in Nickel et al [8] For the test instances thedata is randomly generated and the way for generating data isstated in our manuscript The datasets generated during thecurrent study are available from the corresponding author onreasonable request
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This work was supported by the National Natural Sci-ence Foundation of China (NSFC) under Grants 7142800271531011 71571134 and 71771048 This work was also sup-ported by the Fundamental Research Funds for the CentralUniversities
References
[1] L Brotcorne G Laporte and F Semet ldquoAmbulance loca-tion and relocation modelsrdquo European Journal of OperationalResearch vol 147 no 3 pp 451ndash463 2003
[2] X Li Z Zhao X Zhu and T Wyatt ldquoCovering modelsand optimization techniques for emergency response facilitylocation and planning a reviewrdquo Mathematical Methods ofOperations Research vol 74 no 3 pp 281ndash310 2011
[3] C Toregas R Swain C ReVelle and L Bergman ldquoThe locationof emergency service facilitiesrdquoOperations Research vol 19 no6 pp 1363ndash1373 1971
[4] R Church and C ReVelle ldquoThe maximal covering locationproblemrdquo Papers in Regional Science vol 32 no 1 pp 101ndash1181974
[5] M Gendreau G Laporte and F Semet ldquoSolving an ambulancelocation model by tabu searchrdquo Location Science vol 5 no 2pp 75ndash88 1997
[6] A Ingolfsson S Budge and E Erkut ldquoOptimal ambulancelocation with random delays and travel timesrdquo Health CareManagement Science vol 11 no 3 pp 262ndash274 2008
[7] M Gendreau G Laporte and F Semet ldquoThemaximal expectedcoverage relocation problem for emergency vehiclesrdquo Journal ofthe Operational Research Society vol 57 no 1 pp 22ndash28 2006
[8] S Nickel M Reuter-Oppermann and F Saldanha-da-GamaldquoAmbulance location under stochastic demand A samplingapproachrdquo Operations Research for Health Care vol 8 pp 24ndash32 2016
[9] P Beraldi M E Bruni and D Conforti ldquoDesigning robustemergency medical service via stochastic programmingrdquo Euro-pean Journal of Operational Research vol 158 no 1 pp 183ndash1932004
[10] N Noyan ldquoAlternate risk measures for emergency medicalservice system designrdquo Annals of Operations Research vol 181pp 559ndash589 2010
[11] M R Wagner ldquoStochastic 0-1 linear programming underlimited distributional informationrdquoOperations Research Lettersvol 36 no 2 pp 150ndash156 2008
[12] EDelage andYYe ldquoDistributionally robust optimization undermoment uncertainty with application to data-driven problemsrdquoOperations Research vol 58 no 3 pp 595ndash612 2010
[13] J T Essen J L Hurink S Nickel and M Reuter ldquoModels forambulance planning on the strategic and the tactical levelrdquo BetaResearch School for Operations Management and Logistics p 272013
[14] E Erkut A Ingolfsson and G Erdogan ldquoAmbulance locationfor maximum survivalrdquoNaval Research Logistics (NRL) vol 55no 1 pp 42ndash58 2008
[15] S C Chapman and J A White ldquoProbabilistic formulations ofemergency service facilities location problemsrdquo in Proceedingsof ORSATIMS Conference San Juan Puerto Rico 1974
[16] A Ben-Tal D Den Hertog A De Waegenaere B Melenbergand G Rennen ldquoRobust solutions of optimization problemsaffected by uncertain probabilitiesrdquo Management Science vol59 no 2 pp 341ndash357 2013
[17] MNg ldquoDistribution-free vessel deployment for liner shippingrdquoEuropean Journal of Operational Research vol 238 no 3 pp858ndash862 2014
[18] M Ng ldquoContainer vessel fleet deployment for liner shippingwith stochastic dependencies in shipping demandrdquo Transporta-tion Research Part B Methodological vol 74 pp 79ndash87 2015
[19] R Jiang and Y Guan ldquoData-driven chance constrained stochas-tic programrdquo Mathematical Programming vol 158 no 1-2 SerA pp 291ndash327 2016
[20] C-M Lee and S-L Hsu ldquoThe effect of advertising on thedistribution-free newsboy problemrdquo International Journal ofProduction Economics vol 129 no 1 pp 217ndash224 2011
[21] K Kwon and T Cheong ldquoA minimax distribution-free proce-dure for a newsvendor problem with free shippingrdquo EuropeanJournal of Operational Research vol 232 no 1 pp 234ndash2402014
12 Journal of Advanced Transportation
[22] Y Zhang S Shen and S A Erdogan ldquoDistributionally robustappointment scheduling with moment-based ambiguity setrdquoOperations Research Letters vol 45 no 2 pp 139ndash144 2017
[23] Y Zhang S Shen and S A Erdogan ldquoSolving 0-1semidefinite programs for distributionally robust allocationof surgery blocksrdquo Social Science Electronic Publishing 2017httpsssrncomabstract=2997524
[24] F Zheng J He F Chu and M Liu ldquoA new distribution-freemodel for disassembly line balancing problem with stochastictask processing timesrdquo International Journal of ProductionResearch pp 1ndash13 2018
[25] J Cheng E Delage and A Lisser ldquoDistributionally robuststochastic knapsack problemrdquo SIAM Journal on Optimizationvol 24 no 3 pp 1485ndash1506 2014
[26] M Liu D Yang and F Hao ldquoOptimization for the locationsof ambulances under two-stage life rescue in the emergencymedical service a case study in Shanghai ChinardquoMathematicalProblems in Engineering vol 2017 Article ID 1830480 14 pages2017
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
10 Journal of Advanced Transportation
Table6Com
putatio
nalresultson
insta
nces
with
datageneratedfro
mno
rmaldistr
ibution
set ( |119868 |
|119869 |)120572=
085120572=
09120572=
095Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2Samplingapproach
MIP-D
F1MIP-D
F2CT
Obj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
CTObj
Sel()
Sel()
Sel()
1(26)
27
907294
9964
9117
27
947844
9970
9202
2558
106
8422
10000
9208
2(412
)46
170
7961
9921
9283
48
183
8328
9982
9283
4508
183
8823
10000
9370
3(618
)68
258
7578
9916
9228
68
280
7798
9971
9417
6803
307
7658
9998
9447
4(824)
93315
8000
9899
9367
96337
8379
9979
9467
9373
368
8438
9999
9579
5(1030)
120
413
7841
9927
9437
124
443
8763
9963
9476
1262
499
8850
9998
9708
6(1236)
155
476
7921
9919
9519
162
504
8124
9982
9582
16021
536
8954
10000
9756
7(14
42)
195
573
7976
9933
9598
196
609
8287
9975
9619
2011
646
8467
9999
9767
8(1648)
232
687
8031
9918
9705
235
727
8115
9965
9650
2404
772
8278
9999
9765
9(1854)
293
692
8086
9898
9759
291
734
8464
9974
9755
29815
796
8617
9998
9835
10(2060)
356
796
7984
9937
9770
358
842
8223
9988
9765
37973
921
8846
10000
9838
11(2266
)421
829
8101
9940
9755
436
877
8211
9986
9786
4514
8959
8497
9999
9789
12(2472)
514
1023
8033
9951
9785
501
1085
8162
9983
9788
52073
1147
8321
9999
9824
13(2678)
592
1023
8138
9916
9885
599
1087
8411
9970
9845
62937
1194
8609
9998
9795
14(2884)
728
1099
8235
9913
9915
722
1168
8435
9987
9806
75053
1270
8567
9997
9834
15(3090)
855
1089
8187
9922
9861
880
1157
8376
9976
9796
89222
1300
8611
9996
9798
16(3296)
1009
1158
8770
9897
9788
966
1231
8845
9984
9823
106288
1359
8997
9999
9848
17(34102)
1154
1299
8434
9925
9849
1137
1377
8601
9980
9798
123869
1493
8798
9996
9877
18(36108)
1407
1411
8321
9927
9866
1339
1495
8301
9965
9822
133807
1601
8567
9999
9834
19(38114
)1702
1506
8206
9915
9805
1672
1598
8977
9978
9845
152999
1694
8998
9999
9847
20(40120)
2091
1635
8528
9918
9889
1916
1737
8805
9976
9837
1890
021833
9012
9999
9791
21(42126)
2843
1622
8550
9934
9913
2432
1718
8886
9974
9813
2399
131815
9008
10000
9815
22(44132)
3300
1736
8449
9940
9859
2795
1843
8786
9977
9756
253106
1948
8997
9999
9826
23(46138)
3741
1934
8133
9911
9876
3143
2050
8876
9981
9785
274583
2173
9014
9999
9858
24(48144)
4531
1961
8245
9909
9846
3515
2090
8344
9993
9819
308089
2211
8661
9999
9883
25(50150)
4715
1924
8111
9929
9888
3959
2041
8279
9996
9837
3510
932158
8527
9999
9827
26(52156)
5223
1911
8096
9898
9857
4301
2025
8251
9997
9757
421747
2146
8497
9999
9914
27(54162)
5994
2095
8387
9908
9775
4835
2226
8497
9974
9780
4614
432359
8695
9999
9837
28(56168)
6406
2111
8411
9906
9798
5336
2240
8558
9991
9820
494113
2371
9249
9999
9816
29(58174)
6897
2176
8012
9916
9791
5948
2311
8211
9984
9765
624835
2445
8531
10000
9851
30(60180)
7436
2391
8111
9913
9766
6558
2533
8305
9968
9795
696116
2682
8575
9999
9799
31(62186)
8118
2282
8119
9928
9806
7069
2422
8417
9973
9873
797768
2564
8667
9999
9846
32(64192)
8895
2519
8223
9919
9757
7857
2675
8631
9982
9769
864813
2827
8976
10000
9905
33(66198)
11235
2533
8563
9928
9811
9162
2688
8547
9979
9828
973089
2843
8657
9999
9846
34(68204)
11587
2534
8249
9926
9847
9549
2686
8798
9938
9856
10523
3041
8999
10000
9877
35(70210)
1164
92715
8098
9895
9872
10345
2880
8602
9998
9869
11272
3052
9021
9999
9897
36(72216)
12224
2825
8403
9958
9770
11165
3002
8657
9999
9839
12456
3175
9113
9999
9864
37(74222)
12978
2830
8357
9937
9795
11850
3005
8496
9998
9894
13458
3176
8827
9999
9905
38(76228)
13426
2844
8267
9966
9759
12795
3019
8604
9999
9797
15155
3121
8978
9999
9853
39(78234)
15003
2955
8303
9956
9867
13816
3132
8588
9996
9885
15041
3310
9016
9999
9912
40(80240)
16495
2978
8257
9954
9826
14896
3160
8495
9997
9837
16563
3349
8957
9999
9933
Average
4619
15855
8174
9925
9742
4077
16828
8457
9981
9743
4350
17938
8732
9999
9794
Journal of Advanced Transportation 11
high service level requirement we recommend MIP-DF1 forsolutions
6 Conclusion
This paper investigates an ambulance location problem withambiguity set of demand in which the probability distribu-tion of the uncertain demand is unknown In such a data-driven environment only the mean and covariance demandsat emergency points are knownThe problem is to determinethe base locations and the number of ambulances at eachbase aiming at minimizing the total cost associated withlocating bases and assigning ambulances We propose adistribution-free model with chance constraints Then twoapproximated MIP formulations with different approxima-tion methods of chance constraints are proposed which arebased on different ambiguity sets of the unknown probabilitydistribution Numerical experiments on benchmarks and120 randomly generated instances are conducted and thecomputational results show that the first approximated MIPformulation is much conservative with overhigh system costwhile the second approximatedMIP formulation ismore costsaving with service level appropriately ensured
In future research wemay consider the following relevantissues First the uncertainty of driving time from a base toan emergency point shall be analysed Besides unexpected orsudden disasters may be taken into consideration Moreoverother uncertainties should be considered such as travel timethe number of available vehicles and so on The stochasticdependence should also further considered
Data Availability
The data of the benchmark instances in our manuscript canbe obtained in Nickel et al [8] For the test instances thedata is randomly generated and the way for generating data isstated in our manuscript The datasets generated during thecurrent study are available from the corresponding author onreasonable request
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This work was supported by the National Natural Sci-ence Foundation of China (NSFC) under Grants 7142800271531011 71571134 and 71771048 This work was also sup-ported by the Fundamental Research Funds for the CentralUniversities
References
[1] L Brotcorne G Laporte and F Semet ldquoAmbulance loca-tion and relocation modelsrdquo European Journal of OperationalResearch vol 147 no 3 pp 451ndash463 2003
[2] X Li Z Zhao X Zhu and T Wyatt ldquoCovering modelsand optimization techniques for emergency response facilitylocation and planning a reviewrdquo Mathematical Methods ofOperations Research vol 74 no 3 pp 281ndash310 2011
[3] C Toregas R Swain C ReVelle and L Bergman ldquoThe locationof emergency service facilitiesrdquoOperations Research vol 19 no6 pp 1363ndash1373 1971
[4] R Church and C ReVelle ldquoThe maximal covering locationproblemrdquo Papers in Regional Science vol 32 no 1 pp 101ndash1181974
[5] M Gendreau G Laporte and F Semet ldquoSolving an ambulancelocation model by tabu searchrdquo Location Science vol 5 no 2pp 75ndash88 1997
[6] A Ingolfsson S Budge and E Erkut ldquoOptimal ambulancelocation with random delays and travel timesrdquo Health CareManagement Science vol 11 no 3 pp 262ndash274 2008
[7] M Gendreau G Laporte and F Semet ldquoThemaximal expectedcoverage relocation problem for emergency vehiclesrdquo Journal ofthe Operational Research Society vol 57 no 1 pp 22ndash28 2006
[8] S Nickel M Reuter-Oppermann and F Saldanha-da-GamaldquoAmbulance location under stochastic demand A samplingapproachrdquo Operations Research for Health Care vol 8 pp 24ndash32 2016
[9] P Beraldi M E Bruni and D Conforti ldquoDesigning robustemergency medical service via stochastic programmingrdquo Euro-pean Journal of Operational Research vol 158 no 1 pp 183ndash1932004
[10] N Noyan ldquoAlternate risk measures for emergency medicalservice system designrdquo Annals of Operations Research vol 181pp 559ndash589 2010
[11] M R Wagner ldquoStochastic 0-1 linear programming underlimited distributional informationrdquoOperations Research Lettersvol 36 no 2 pp 150ndash156 2008
[12] EDelage andYYe ldquoDistributionally robust optimization undermoment uncertainty with application to data-driven problemsrdquoOperations Research vol 58 no 3 pp 595ndash612 2010
[13] J T Essen J L Hurink S Nickel and M Reuter ldquoModels forambulance planning on the strategic and the tactical levelrdquo BetaResearch School for Operations Management and Logistics p 272013
[14] E Erkut A Ingolfsson and G Erdogan ldquoAmbulance locationfor maximum survivalrdquoNaval Research Logistics (NRL) vol 55no 1 pp 42ndash58 2008
[15] S C Chapman and J A White ldquoProbabilistic formulations ofemergency service facilities location problemsrdquo in Proceedingsof ORSATIMS Conference San Juan Puerto Rico 1974
[16] A Ben-Tal D Den Hertog A De Waegenaere B Melenbergand G Rennen ldquoRobust solutions of optimization problemsaffected by uncertain probabilitiesrdquo Management Science vol59 no 2 pp 341ndash357 2013
[17] MNg ldquoDistribution-free vessel deployment for liner shippingrdquoEuropean Journal of Operational Research vol 238 no 3 pp858ndash862 2014
[18] M Ng ldquoContainer vessel fleet deployment for liner shippingwith stochastic dependencies in shipping demandrdquo Transporta-tion Research Part B Methodological vol 74 pp 79ndash87 2015
[19] R Jiang and Y Guan ldquoData-driven chance constrained stochas-tic programrdquo Mathematical Programming vol 158 no 1-2 SerA pp 291ndash327 2016
[20] C-M Lee and S-L Hsu ldquoThe effect of advertising on thedistribution-free newsboy problemrdquo International Journal ofProduction Economics vol 129 no 1 pp 217ndash224 2011
[21] K Kwon and T Cheong ldquoA minimax distribution-free proce-dure for a newsvendor problem with free shippingrdquo EuropeanJournal of Operational Research vol 232 no 1 pp 234ndash2402014
12 Journal of Advanced Transportation
[22] Y Zhang S Shen and S A Erdogan ldquoDistributionally robustappointment scheduling with moment-based ambiguity setrdquoOperations Research Letters vol 45 no 2 pp 139ndash144 2017
[23] Y Zhang S Shen and S A Erdogan ldquoSolving 0-1semidefinite programs for distributionally robust allocationof surgery blocksrdquo Social Science Electronic Publishing 2017httpsssrncomabstract=2997524
[24] F Zheng J He F Chu and M Liu ldquoA new distribution-freemodel for disassembly line balancing problem with stochastictask processing timesrdquo International Journal of ProductionResearch pp 1ndash13 2018
[25] J Cheng E Delage and A Lisser ldquoDistributionally robuststochastic knapsack problemrdquo SIAM Journal on Optimizationvol 24 no 3 pp 1485ndash1506 2014
[26] M Liu D Yang and F Hao ldquoOptimization for the locationsof ambulances under two-stage life rescue in the emergencymedical service a case study in Shanghai ChinardquoMathematicalProblems in Engineering vol 2017 Article ID 1830480 14 pages2017
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
Journal of Advanced Transportation 11
high service level requirement we recommend MIP-DF1 forsolutions
6 Conclusion
This paper investigates an ambulance location problem withambiguity set of demand in which the probability distribu-tion of the uncertain demand is unknown In such a data-driven environment only the mean and covariance demandsat emergency points are knownThe problem is to determinethe base locations and the number of ambulances at eachbase aiming at minimizing the total cost associated withlocating bases and assigning ambulances We propose adistribution-free model with chance constraints Then twoapproximated MIP formulations with different approxima-tion methods of chance constraints are proposed which arebased on different ambiguity sets of the unknown probabilitydistribution Numerical experiments on benchmarks and120 randomly generated instances are conducted and thecomputational results show that the first approximated MIPformulation is much conservative with overhigh system costwhile the second approximatedMIP formulation ismore costsaving with service level appropriately ensured
In future research wemay consider the following relevantissues First the uncertainty of driving time from a base toan emergency point shall be analysed Besides unexpected orsudden disasters may be taken into consideration Moreoverother uncertainties should be considered such as travel timethe number of available vehicles and so on The stochasticdependence should also further considered
Data Availability
The data of the benchmark instances in our manuscript canbe obtained in Nickel et al [8] For the test instances thedata is randomly generated and the way for generating data isstated in our manuscript The datasets generated during thecurrent study are available from the corresponding author onreasonable request
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This work was supported by the National Natural Sci-ence Foundation of China (NSFC) under Grants 7142800271531011 71571134 and 71771048 This work was also sup-ported by the Fundamental Research Funds for the CentralUniversities
References
[1] L Brotcorne G Laporte and F Semet ldquoAmbulance loca-tion and relocation modelsrdquo European Journal of OperationalResearch vol 147 no 3 pp 451ndash463 2003
[2] X Li Z Zhao X Zhu and T Wyatt ldquoCovering modelsand optimization techniques for emergency response facilitylocation and planning a reviewrdquo Mathematical Methods ofOperations Research vol 74 no 3 pp 281ndash310 2011
[3] C Toregas R Swain C ReVelle and L Bergman ldquoThe locationof emergency service facilitiesrdquoOperations Research vol 19 no6 pp 1363ndash1373 1971
[4] R Church and C ReVelle ldquoThe maximal covering locationproblemrdquo Papers in Regional Science vol 32 no 1 pp 101ndash1181974
[5] M Gendreau G Laporte and F Semet ldquoSolving an ambulancelocation model by tabu searchrdquo Location Science vol 5 no 2pp 75ndash88 1997
[6] A Ingolfsson S Budge and E Erkut ldquoOptimal ambulancelocation with random delays and travel timesrdquo Health CareManagement Science vol 11 no 3 pp 262ndash274 2008
[7] M Gendreau G Laporte and F Semet ldquoThemaximal expectedcoverage relocation problem for emergency vehiclesrdquo Journal ofthe Operational Research Society vol 57 no 1 pp 22ndash28 2006
[8] S Nickel M Reuter-Oppermann and F Saldanha-da-GamaldquoAmbulance location under stochastic demand A samplingapproachrdquo Operations Research for Health Care vol 8 pp 24ndash32 2016
[9] P Beraldi M E Bruni and D Conforti ldquoDesigning robustemergency medical service via stochastic programmingrdquo Euro-pean Journal of Operational Research vol 158 no 1 pp 183ndash1932004
[10] N Noyan ldquoAlternate risk measures for emergency medicalservice system designrdquo Annals of Operations Research vol 181pp 559ndash589 2010
[11] M R Wagner ldquoStochastic 0-1 linear programming underlimited distributional informationrdquoOperations Research Lettersvol 36 no 2 pp 150ndash156 2008
[12] EDelage andYYe ldquoDistributionally robust optimization undermoment uncertainty with application to data-driven problemsrdquoOperations Research vol 58 no 3 pp 595ndash612 2010
[13] J T Essen J L Hurink S Nickel and M Reuter ldquoModels forambulance planning on the strategic and the tactical levelrdquo BetaResearch School for Operations Management and Logistics p 272013
[14] E Erkut A Ingolfsson and G Erdogan ldquoAmbulance locationfor maximum survivalrdquoNaval Research Logistics (NRL) vol 55no 1 pp 42ndash58 2008
[15] S C Chapman and J A White ldquoProbabilistic formulations ofemergency service facilities location problemsrdquo in Proceedingsof ORSATIMS Conference San Juan Puerto Rico 1974
[16] A Ben-Tal D Den Hertog A De Waegenaere B Melenbergand G Rennen ldquoRobust solutions of optimization problemsaffected by uncertain probabilitiesrdquo Management Science vol59 no 2 pp 341ndash357 2013
[17] MNg ldquoDistribution-free vessel deployment for liner shippingrdquoEuropean Journal of Operational Research vol 238 no 3 pp858ndash862 2014
[18] M Ng ldquoContainer vessel fleet deployment for liner shippingwith stochastic dependencies in shipping demandrdquo Transporta-tion Research Part B Methodological vol 74 pp 79ndash87 2015
[19] R Jiang and Y Guan ldquoData-driven chance constrained stochas-tic programrdquo Mathematical Programming vol 158 no 1-2 SerA pp 291ndash327 2016
[20] C-M Lee and S-L Hsu ldquoThe effect of advertising on thedistribution-free newsboy problemrdquo International Journal ofProduction Economics vol 129 no 1 pp 217ndash224 2011
[21] K Kwon and T Cheong ldquoA minimax distribution-free proce-dure for a newsvendor problem with free shippingrdquo EuropeanJournal of Operational Research vol 232 no 1 pp 234ndash2402014
12 Journal of Advanced Transportation
[22] Y Zhang S Shen and S A Erdogan ldquoDistributionally robustappointment scheduling with moment-based ambiguity setrdquoOperations Research Letters vol 45 no 2 pp 139ndash144 2017
[23] Y Zhang S Shen and S A Erdogan ldquoSolving 0-1semidefinite programs for distributionally robust allocationof surgery blocksrdquo Social Science Electronic Publishing 2017httpsssrncomabstract=2997524
[24] F Zheng J He F Chu and M Liu ldquoA new distribution-freemodel for disassembly line balancing problem with stochastictask processing timesrdquo International Journal of ProductionResearch pp 1ndash13 2018
[25] J Cheng E Delage and A Lisser ldquoDistributionally robuststochastic knapsack problemrdquo SIAM Journal on Optimizationvol 24 no 3 pp 1485ndash1506 2014
[26] M Liu D Yang and F Hao ldquoOptimization for the locationsof ambulances under two-stage life rescue in the emergencymedical service a case study in Shanghai ChinardquoMathematicalProblems in Engineering vol 2017 Article ID 1830480 14 pages2017
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
12 Journal of Advanced Transportation
[22] Y Zhang S Shen and S A Erdogan ldquoDistributionally robustappointment scheduling with moment-based ambiguity setrdquoOperations Research Letters vol 45 no 2 pp 139ndash144 2017
[23] Y Zhang S Shen and S A Erdogan ldquoSolving 0-1semidefinite programs for distributionally robust allocationof surgery blocksrdquo Social Science Electronic Publishing 2017httpsssrncomabstract=2997524
[24] F Zheng J He F Chu and M Liu ldquoA new distribution-freemodel for disassembly line balancing problem with stochastictask processing timesrdquo International Journal of ProductionResearch pp 1ndash13 2018
[25] J Cheng E Delage and A Lisser ldquoDistributionally robuststochastic knapsack problemrdquo SIAM Journal on Optimizationvol 24 no 3 pp 1485ndash1506 2014
[26] M Liu D Yang and F Hao ldquoOptimization for the locationsof ambulances under two-stage life rescue in the emergencymedical service a case study in Shanghai ChinardquoMathematicalProblems in Engineering vol 2017 Article ID 1830480 14 pages2017
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom