assignment models - didatticaweb
TRANSCRIPT
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Teoria dei Sistemi di Trasporto 1 + 2 (6/9 CFU)
A. A. 2020 - 2021
Assignment models
(elements)
prof. ing. Antonio Comi
Department of Enterprise Engineering
University of Rome Tor Vergata
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Transportation system model structure
2
TRANSPORTATION SYSTEM MODEL
TRANSPORT
FACILITIES AND
SERVICES
MODELS OF ACTIVITY
LOCATION AND
LEVEL
ACTIVITY
SYSTEM
Level of Service attributes
(times, costs)
ASSIGNMENT
MODELS
DEMAND MODEL
O/D MATRICES
Performance functions
SUPPLY MODEL
Transportation
networks
Flows
Evaluation
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Assignment models
3
They simulate the interaction of demand and supply on a
transportation network allowing:
➢ calculation of user flows
➢ calculation of performance measures for each supply element
(network link)
resulting from origin–destination (O-D) demand flows, path choice
behaviour, and the mutual interactions between supply and demand.
Assignment model outputs are inputs required for design and/or
evaluation of transportation projects.
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Classification criteria
The system state simulated through assignment models depends on
assumptions about user behaviour (demand functions, path choice,
available information) and the approach used for representing supply–
demand interactions.
✓ Supply-demand interactions➢ Equilibrium assignment
➢ Process assignment
✓ Cost functions➢ Congested networks
➢ Not-Congested networks
✓ Path choice behaviour➢ Deterministic model
➢ Stochastic model
4
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Classification [1/3]
Demand-supply interaction
The user equilibrium (UE) represents equilibrium configurations of
the system, that is, configurations in which demand, path, and link
flows are consistent with the costs that they produce in the network.
From a mathematical point of view, equilibrium assignment can be
defined as the problem of finding a flow vector that reproduces itself
based on the correspondence defined by the supply and demand
models (fixed point models, or else with variational inequality or
optimization models).
5
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Classification [2/3]
Demand-supply interaction
The alternative approach for representing supply–demand
interactions leads to between-period (or day-to-day) dynamic process
(DP) assignment models.
It is assumed that the system evolves over time
✓ through possibly different feasible states,
✓ as a result of changes in the number of users undertaking trips, path choices,
supply performance, and so on.
One of the mechanisms that drives the changes from one state to
another is the dependency between flows and costs.
6
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Classification [3/3]
Demand-supply interaction
In a given reference period, the system state – defined by the demand,
path, and link flows and the corresponding costs – may be internally
inconsistent, and this may cause a change towards a different state in
the following reference periods.
Dynamic process assignment models explicitly simulate the evolution
of the system state based on the mechanisms underlying path choice
and information acquisition, which in turn determine user choices in
successive reference periods.
By analogy, equilibrium assignment could be termed within-day
static assignment.
7
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Assignment models classification
8
DNL = Deterministic Network Loading
AoN = All or Nothing
SNL = Stochastic Network Loading
DUE = Deterministic User Equilibrium
SUE = Stochastic User Equilibrium
PATH CHIOCE MODEL
DETERMINISTIC STOCHASTIC
NOT CONGESTED
NETWORKDNL (AoN) SNL
CONGESTEDEQUILIBRIUM DUE SUE
NETWORK DYNAMIC
PROCESSDDP SDP
DDP = Deterministic Dynamic Process
SDP = Stochastic Dynamic Process
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Assignment models classification
9
ASSIGNMENT MODELS Path choice model
Deterministic Stochastic
c = cost DNL (AoN) SNL
c = c(f) equilibrium DUE SUE
DNL = Deterministic Network Loading
AoN = All or Nothing
SNL = Stochastic Network Loading
DUE = Deterministic User Equilibrium
SUE = Stochastic User Equilibrium
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Assignment with inelastic and elastic
demandAssignment models with inelastic/elastic demand assume that O-D
demand flows are independent/dependent of changes in network
costs that may occur as a result of congestion.
Elastic demand
demand flows vary with congestion costs; demand flows are therefore a function
of path costs resulting from congestion, as well as of activity system attributes.
Depending on the modelling context, demand might be assumed variable in
certain choice dimensions only. For example, it might be assumed that the total
O-D matrix is cost-independent (meaning that frequency and destination choices
are not influenced by cost variations), but that mode choice is affected by the
relative costs of the available modes; in this way, multimode assignment models
are obtained. Obviously, from a practical viewpoint, demand elasticity is relevant
only for congested networks where costs depend on flows.
10
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Schematic representation of assignment
models
11
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Fields of Application
✓ Assignment models as estimators of the present state (monitoring)
of the transportation systemThe assignment model is applied to estimate other quantities that would be too.
✓Assignment models for simulating the effects of modifications to
the transportation system.
Assignment models are used to estimate the changes in relevant network
variables due to changes in supply and/or demand.
✓Assignment models for the estimation (or update) of travel demand.
Assignment models are seeing increasing application for the estimation of O-D
demand flows and/or for the calibration of demand models.
12
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Definitions and assumptionsRelationship between link costs and path costs
13
Let:
o be a origin centroid node;
d be a destination centroid node;
od be an origin–destination pair;
Iod be the set of paths for O-D pair od; each path k is uniquely
associated with one and only one O-D pair od;
A be the overall link–path incidence matrix;
c be the link cost vector, with entries cl;
C be the overall path cost vector, with entries Cod (the path costs
for users of O-D pair od);
C = AT c
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Definitions and assumptions
101101)4,3(5
010010)4,2(4
001001)3,2(3
000100)3,1(2
000011)2,1(1
654321
=A
14
+
+
+
++
=
===
=
34
24
3423
3413
2412
342312
5
4
3
2
1
6
5
4
3
2
1
10000
01000
10100
10010
01001
10101
cc
cc
cc
cc
ccc
c
c
c
c
c
cAC
C
C
C
C
C
C
C TADD
G (N,L)
N (1, 2, 3,4)
L (1, 2),(1, 3),(2, 3),(2, 4),(3, 4)
Origins1, 2, 3 Destination4
6
Relationship between link costs and path costs C = AT c
1
2
3
4
GRAPHPATHS
3
4
4
2
3
41
2
3
41
2
451
2
42
3
1 43
1
2
3
4
5
=
=
1
2
3
1
2
10000
01000
10100
10010
01001
10101
1
2
4
2
4
6cAC TC = AT c
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Definitions and assumptions
15
For not congested networks:
c = cost
For congested networks:
c = c(f)
C = AT c(f)
Let:
f be the link flow vector, with entries, with entries fl;
F be the overall vector of path flows, consisting of the vectors of
path flows Fod related to each pair od;
will be:
f = A F
0,0
0,5
1,0
1,5
2,0
2,5
3,0
180 540 900 1260 1620 1980 2340 2700
flusso
co
sto
g=2
g=3
g=4
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Definitions and assumptions
101101)4,3(5
010010)4,2(4
001001)3,2(3
000100)3,1(2
000011)2,1(1
654321
=A
16
=
=
800
1321
179
867
117
16
101101
010010
001001
000100
000011
1861
1439
195
867
133
FAf
G (N,L)
N (1, 2, 3, 4)
L (1, 2),(1, 3),(2, 3),(2, 4),(3, 4)
Origins1, 2, 3 Destination 4
1
2
3
GRAPHPATHS
1
2
3
4
5
Relationship between link flows and path flows f = A F
6
4
3
4
4
2
3
41
2
3
41
2
451
2
42
3
1 43
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Path choice model
17
d whose components are the demand values dod for each O-D pair.
Path choice behaviour
Random Utility Model
p[k/od] = Prob[Vk - Vj j - k j Є Iod] od,k
Deterministic utility model Stochastic utility model
0=ε 0ε
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Path choice model
18
pod = pod(Cod) od
P = P(C)
pod be the vector of path choice probabilities for users of O-D pair
od;
P the matrix of path choice probabilities for users of O-D pair
od, with a column for each O-D pair od and a row for each
path k;
the generic entry is p[k/od] if exist path k between od, 0
otherwise.
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
1-4 2-4 3-4
1 0.016 0 0
2 0.117 0 0
3 0.867 0 0
4 0 0.119 0
5 0 0.881 0
6 0 0 1.000
Paths
Path choice matrix
19
( )( )
od,k
od,nn
exp Cp k / od
exp C
1
−=
−
=
14 24 34
0.0160.119
= 0.117 ; p = ; = 1.0000.881
0.867
p p
O-D
HP: stochastic path choice model.
6
4
2
4
2
1
=
C
P
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Path flow
20
Fk = dod p[k/od]
F = P (C) d
where
Fk = flow on path k
dod = demand flow on od pair
p[k/od] = path k (from o to d) choice probability
F = path flow vector
P = matrix of path choice probabilities
C = path cost vector
d = demand vector
A = link-path incidence matrix
c = link cost vector
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Assignment model
21
Fk = dod p[k/od]
F = P (C) d
f = A F = A P (C) d
F = path flow vector;
P = matrix of path choice probabilities;
C = path cost vector;
d = demand vector;
f = flow cost vector;
A = link-path incidence matrix;
c = link cost vector.
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Uncongested network assignment models
22
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
SNL assignment modelsPath flows
23
( )( )
1α Cα
Cαp
nodn
kodkod =
−
−=
= ,
,,
exp
exp
1
2
4
2
4
6
C
000.1p ;881.0
119.0p ;
867.0
117.0
016.0
p 342414 =
=
=
Percorsi
Coppie O-D 1-4 2-4 3-4
1 0.016 0 0
2 0.117 0 0
3 0.867 0 0
4 0 0.119 0
5 0 0.881 0
6 0 0 1.000
=
=
800
1500
1000
000.100
0881.00
0119.00
00867.0
00117.0
00016.0
800
1321
179
867
117
16dPF
HP: stochastic path choice model
HP: constant link costs
( )( )
od,k
od,nn
exp Cp k / od
exp C
1
−=
−
=
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
SNL assignment modelsPath flows
24
fNL = fNL(c; d) = A P(AT c) d c
16133 1 1 0 0 0 0
117867 0 0 1 0 0 0
867195 1 0 0 1 0 0
1791438 0 1 0 0 1 0
13211862 1 0 1 1 0 1
800
=
f = A · F
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
SNL assignment models
25
d1-4 = 1000
16
1616
3
2
1 4
2
1 4
117 117
3
1 4
867 867
d2-4 = 1500
1
3
2
4179
179
2
4
1321
d3-4 = 800
3
2
1 4
800
Link flows
3
2
1 4
1862867
195
1438133
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
DNL (AoN) assignment modelsPath flow
26
1 ;1
0 ;
1
0
0
1
2
4
2
4
6
342414 =
=
=
= pppC
1-4 2-4 3-4
1 0 0 0
2 0 0 0
3 1 0 0
4 0 0 0
5 0 1 0
6 0 0 1
PathsO-D
=
=
800
1500
1000
100
010
000
001
000
000
800
1500
0
1000
0
0dPF
HP: deterministic path choice model
HP: constant link costs
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
DNL (AoN) assignment modelsLink flows
27
fNL = fNL(c; d) = A P(AT c) d c
=
800
1500
0
1000
0
0
101101
010010
001001
000100
000011
1800
1500
0
1000
0
f = A · F
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
DNL (AoN) assignment modelsPath flow
28
d1-4 = 1000
0
00
3
2
1 4
2
1 4
0 0
3
1 4
1000 1000
d2-4 = 1500
1
3
2
40
0
2
4
1500
d3-4 = 800
3
2
1 4
800
Link flows
3
2
1 4
18001000
0
15000
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Assignment models classification
29
ASSIGNMENT MODELS Path choice
Deterministic Stochastic
c = cost DNL (AoN) SNL
c = c(f) equilibrium DUE SUE
DNL = Deterministic Network Loading
AoN = All or Nothing
SNL = Stochastic Network Loading
DUE = Deterministic User Equilibrium
SUE = Stochastic User Equilibrium
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Assignment models classificationEquilibrium
30
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Assignment models classificationEquilibrium
31
c = c (f)
f = A F
c = c (A F)
circular dependence
between flows and
costs
F* = P [AT c (A F*)] d
f* = A P [AT c (f*)] d
The state, or configuration, of a transport network in a specific
reference time (for example in the morning rush hour of a day),
can be defined by the path flows vector F, from this it is
possible going back to the link flows vector f and then to the
link costs and path costs vector.
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Equilibrium problem: notionsExample
32
od
1
2
veh100
hodd =
0
10
20
30
40
50
60
70
80
90
100
0 20 40 60 80 100f1
c1
0
10
20
30
40
50
60
70
80
90
100
0 20 40 60 80 100f2
c2
( ) ffc 111 5.030 +=
( ) ffc 222 5.050 +=
Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)Prof. Ing. Antonio Comi Teoria dei Sistemi di Trasporto 1 + 2 (9 CFU)
Equilibrium problem: notionsExample
33
55107590
500805030100
22
11
22
11
=→==→=
−−−−−−−−−−−=→=
=+=→=
C fC f
C fCf
6530
6570
22
11
=→=
=→=
Cf
Cf
c
0
10
20
30
40
50
60
70
80
90
100
0 20 40 60 80 100
c1
c2
f1
40 20 06080100
f2
70
30
c