w{cdma network design - lyle school of engineeringlyle.smu.edu/~olinick/cdma2.pdf · 2005-01-15 ·...
TRANSCRIPT
Technical Report 03-EMIS-02
W–CDMA Network Design∗
Qibin Cai1
Joakim Kalvenes2
Jeffery Kennington1
Eli Olinick1
1 {qcai,jlk,olinick}@engr.smu.edu
School of Engineering
Southern Methodist University
Dallas, TX 75275-0122
Edwin L. Cox School of Business
Southern Methodist University
Dallas, TX 75275-0333
December 2003
∗This research was supported in part by the Office of Naval Research Award Number N00014-96-1-0315.
Abstract
In this investigation, the W–CDMA network design problem is modeled as a discrete
optimization problem that maximizes revenue net the cost of constructing base stations,
mobile telephone switching offices, and the backbone network to connect base stations
through mobile telephone switching offices to the public switched telephone network.
The formulation results in a very large scale integer programming problem with up to
18,000 integer variables and 20,000 constraints. To solve this large-scale integer pro-
gramming problem, we develop a pair of models, one for the upper bound and one for
the lower bound. The upper bound model relaxes integrality on some of the variables
while the lower bound model uses a 5% optimality gap to achieve early termination.
Additionally, we develop a heuristic procedure that can solve the largest problem in-
stances very quickly with a small optimality gap. To demonstrate the efficiency of the
proposed solution methods, problem instances were solved with five candidate mobile
telephone switching offices servicing some 11,000 simultaneous cellular phone sessions
on a network with up to 160 base stations. In all instances, solutions guaranteed to be
within 5% of optimality were obtained in less than an hour of CPU time.
1 Introduction
Third generation mobile communication systems currently under development promise to
provide its subscribers with high-speed data services at rates up to a hundred times that of
second generation voice channels. There are two accepted major standards for third genera-
tion mobile systems (W–CDMA and CDMA2000, respectively), both of which are based on
code division multiple access (CDMA) technology. This manuscript presents a comprehen-
sive model of the wideband CDMA network design problem. Model features include mobile
switching office (MTSO) and base station (tower) site selection, backbone network design,
and customer service assignment to selected towers. CDMA network design problems differ
considerably from other wireless network design problems in that channel allocation is not
an explicit issue. In each cell, all of the bandwidth available to the service provider can be
used. The features in CDMA making this possible are stringent power control of all system
devices (including user handsets) and the use of orthogonal codes to ensure minimal inter-
ference between simultaneous sessions. Instead, however, the network design must take into
consideration the system-wide interference generated by the mobile users in the service area.
Previous work on CDMA system design has focused on base station location and cus-
tomer assignment. Galota et al. (2001) proposed a profit maximization model for base station
location and customer service assignment based on a limited interference model. Similarly,
Mathar and Schmeink (2001) developed a budget-constrained system capacity maximization
model, in which the interference model accounted for base stations utilized instead of the
number of customers serviced by each respective base station. Amaldi et al. (2001a) pro-
vided a cost minimization model that explicitly considers the signal-to-interference conditions
generated by the base station location and customer service assignment choices by means of
a penalty term in the objective function. Building upon this work, Kalvenes et al. (2003)
developed a profit maximization model in which the signal-to-interference requirements are
enforced as constraints in the mathematical programing model.
In another stream of work, researchers have modeled the selection of MTSOs and the
assignment of base stations to MTSOs. Merchant and Sengupta (1995) developed a cost
minimization model that includes base station to MTSO wiring cost and handoff cost for
given traffic volume at the base stations. The same concept was refined by Li et al. (1997).
Neither investigation includes cost for connecting the MTSOs to one another or to the public
switched telephone network (PSTN).
This investigation extends the basic ideas presented in Kalvenes et al. (2003) and pro-
vides a comprehensive model of W–CDMA network design, including the selection of base
stations and MTSO locations, the assignment of customer locations to base stations, and
the design of a spanning tree to connect the base stations, MTSOs and the PSTN gateway.
The selection of base stations and MTSOs combined with the design of the spanning tree is
equivalent to a Steiner tree problem (see, for instance, Beasley (1989)).
The contributions of this work are several. First, we provide the first comprehensive
discrete optimization model for the W–CDMA network design problem. The model maxi-
mizes the net revenue of service provisioning to mobile subscribers and takes into account
the cost of tower construction, MTSO location, tower to MTSO connection, and MTSO to
PSTN gateway connection. When selecting base station locations, the revenue potential of
each tower is balanced with its cost of installation and operation while simultaneously ensur-
ing sufficient quality of service. The selected base stations are then connected to a network
2
of MTSOs that is generated based on the cost of MTSO location and the cost of wiring
from the towers to the MTSO locations. Second, we develop a unique solution strategy that
involves the application of discrete models to obtain both upper bounds and good feasible
solutions. The solution procedures exploit the problem structure through the addition of
valid inequalities to the model formulation. Third, we develop a new heuristic procedure
that substantially reduces the computational burden for the most difficult problem instances
while resulting in only small reductions in objective function value. Finally, we demonstrate
the efficiency of our solution procedures by solving 40 randomly generated test cases and
seven test cases from the North Dallas area, comparing the three solution procedures pro-
posed in this manuscipt. The very reasonable computational times and the quality of the
obtained solutions are very encouraging. The software implementation of both our exact
solution procedure and our heuristic procedures have been placed in the public domain at
http://www.engr.smu.edu/˜jlk/publications/publications.htm so that both practitioners and
other research groups can experiment with our software and compare computational results.
2 W–CDMA Network Design Model
Our model differs from previous work in that it simultaneously selects base station and MTSO
locations, connects the towers and MTSOs to the PSTN access point, and provides service
assignment of customer locations to base stations based on a realistic interference model.
Thus, this is the first comprehensive planning model for W–CDMA network design.
3
2.1 Sets Used in the Model
Let L denote the set of candidate locations for tower construction. There is a set of subscriber
locations, M . The set Cm ⊂ L is the set of candidate towers that are able to service customers
in location m ∈ M , as determined by the maximum handset transmission power. For every
` ∈ L, P` ⊂ M is the set of customer locations that can be serviced by tower `. Each selected
tower location will be connected to a mobile telephone switching office (MTSO). The set
of candidate MTSO locations is K. In addition, there is a gateway to the public switched
telephone network which is labeled location 0. The union of the PSTN gateway and the set
K is denoted K0.
2.2 Constants Used in the Model
The demand for service in customer area m ∈ M is denoted by dm. This value is the
number of channel equivalents1 required to service the population in the area at an acceptable
service level (call blocking rate). Let r denote the annual revenue (in $) generated by each
channel equivalent utilized in a customer area. The cost (amortized annually) of building
and operating a tower at location ` ∈ L and connecting it to the backbone network is given
by the parameter a`. Operating cost includes the cost of transmission power, marketing,
accounting, customer aquisition and retention, and any other cost that is contingent upon
operating a tower. When a subscriber in location m is serviced by tower `, the subscriber’s
handset must transmit with sufficient power so that the tower receives it at the target power
level Ptarget. Due to attenuation, the signal transmitted weakens over the path from the
1CDMA does not utilize channles to allocate bandwidth to sessions, but an equivalent maximum trans-mission bitrate is allocated to sessions through the use of orthogonal spreading codes.
4
handset to the tower based on the relative location of the origin and destination (depending
on distance, topography, local conditions, etc.). The attenuation factor from subscriber
location m to tower location ` is given by the parameter gm`. To ensure proper received
power, Ptarget, at the tower location, the handset will transmit with power level Ptarget/gm`. At
each tower location, signals are received from many subscriber handsets in the surrounding
neighborhood. In order for the voice packets to be processed with a reasonable error rate,
the signal to interference ratio for any active session must be more than the threshold value
SIRmin. The selected towers will be connected to an MTSO. The MTSOs are limited in
the number of base stations they can service. This limit is given by the parameter α. The
annualized cost of providing a link between tower location ` ∈ L and MTSO hub location
k ∈ K is given by c`k, while hjk is the annualized cost of providing a link from hub location
j ∈ K to hub location k ∈ K0. Finally, bk is the annualized cost of locating an MTSO in
location k ∈ K.
2.3 Decision Variables Used in the Model
The decision variables in this model include general integer and binary variables. The decision
to build a tower at a candidate location is represented by variable y`, which is one if a tower is
built at location l ∈ L; and zero, otherwise. The integer variable xm` represents the capacity
assignment (in channel equivalents) to tower ` ∈ L for servicing of customers in location
m ∈ M . In other words,∑
m∈M xm` represents the instantaneous communication capacity
of tower location ` ∈ L. The variables are related so that xm` ≥ 1 only if y` = 1, that is,
customers in location m can be assigned to tower ` for service only if tower ` is built. If an
MTSO is established in location k ∈ K0, the variable zk is one; and zero, otherwise. Each
5
tower must be linked to an MTSO. If tower ` ∈ L is connected to MTSO k ∈ K, then s`k
is one; otherwise, it is zero. Finally, each MTSO location must have a path to the PSTN
gateway. We use a flow formulation to create a path from every selected MTSO location
to the PSTN gateway. The integer variable ujk denotes the units of traffic flow on the link
between MTSO location j ∈ K and MTSO location k ∈ K0. If there is any flow from MTSO
location j ∈ K to MTSO location k ∈ K0, then a link between j and k has to be established.
The variable wjk is one if a link is established between locations j ∈ K and k ∈ K0; and zero,
otherwise.
2.4 Quality of Service Constraint
In spread-spectrum system design, it is customary to express quality of a communication
link in terms of a signal-to-interference ratio. A derivation of the signal-to-interference ratio
based on the available bandwidth and the link quality requirements can be found in Kalvenes
et al. (2003).
The total received power at tower location `, P TOT
` , from all mobile users in the service
area is given by
P TOT
` = Ptarget
∑
m∈M
∑
j∈Cm
gm`
gmj
xmj. (1)
In this expression, the signal level from customers assigned to tower ` is Ptarget, while it
is Ptargetgm`/gmj from customers assigned to some other tower j. From a single customer’s
perspective, the signals from other customers represent interference. Thus, for each session
assigned to tower `, P TOT
` − Ptarget represents interference, while Ptarget is the signal strength
associated with the session (Amaldi et al. 2001b). Consequently, a quality of service con-
6
straint based on the threshold signal to interference ratio for each session assigned to tower
` is given by
Ptarget
P TOT
` − Ptarget
≥ SIRmin, (2)
provided that tower ` is constructed. Since the tower is built only if y` = 1, this constraint
can be written as follows:
∑
m∈M
∑
j∈Cm
gm`
gmj
xmj ≤ 1 +1
SIRmin
+ (1 − y`)β` ∀` ∈ L, (3)
where β` =∑
m∈M dm
{
maxm∈Cm\{`}
(gm`
gmj
) }
and maxj∈Cm\{`}
(gm`
gmj
)
= 0 if Cm \ {`} = ∅.
The second term on the right-hand side is zero when a tower is built (y` = 1), so that the
signal-to-interference requirement must be met at tower `. When y` = 0, the right-hand side
is so large that the constraint is automatically satisfied.
2.5 Mathematical Formulation
The base station and MTSO location with backbone network design problem is formulated
as follows.
max r∑
m∈M
∑
`∈Cm
xm`
︸ ︷︷ ︸
−∑
`∈L
a`y`
︸ ︷︷ ︸
−∑
k∈K
bkzk
︸ ︷︷ ︸
−∑
`∈L
∑
k∈K
c`ks`k
︸ ︷︷ ︸
−∑
j∈K
∑
k∈K0\{j}
hjkwjk
︸ ︷︷ ︸
.
Subscriberrevenue
Towercost
MTSOcost
Connectioncost
Backbonecost
(4)
There are 16 sets of constraints that define the model. The first set ensures that
7
customers can be serviced only if there are towers that cover the demand area.
xm` ≤ dmy` ∀m ∈ M, ` ∈ Cm. (5)
The next set of constraints ensures that one cannot serve more customers in a location than
there is demand for service.
∑
`∈Cm
xm` ≤ dm ∀m ∈ M. (6)
The next set of constraints enforce the quality of service restrictions on received signal quality
at the towers.
∑
m∈M
∑
j∈Cm
gm`
gmj
xmj ≤ 1 +1
SIRmin+ (1 − y`)β` ∀` ∈ L. (7)
The following two sets of constraints ensure that each base station is connected to an
MTSO and that an MTSO is installed if there is a base station connected to it.
∑
k∈K
s`k = y` ∀` ∈ L, (8)
s`k ≤ zk ∀` ∈ L, ∀k ∈ K. (9)
The capacity constraint on the number of base stations that can be serviced by an
8
MTSO is given by the set of constraints
∑
`∈L
s`k ≤ αzk ∀k ∈ K. (10)
The selected MTSO locations must be connected to the public switched telephone
network gateway either directly or indirectly via another selected MTSO location. We use
a flow formulation that results in a spanning tree with the PSTN gateway as its root. The
first set of constraints provides a link between MTSOs j ∈ K and k ∈ K0 if there is any
flow over the link. The second set of constraints ensures that there can be traffic flow from
MTSO location j to MTSO location k (or the PSTN gateway) only if MTSO location k is
constructed. If zk = 1, then one unit of flow will be generated at MTSO location k. The third
set of constraints represents flow conservation where the flow-out minus the flow-in equals
the flow generated at each MTSO location k. The last constraint ensures that the flow into
the PSTN gateway equals the number of MTSOs selected.
ujk ≤ |K|wjk ∀j ∈ K, k ∈ K0 \ {j}, (11)
ujk ≤ |K|zk ∀j ∈ K, k ∈ K0 \ {j}, (12)
∑
j∈K\{k}
(ukj − ujk) + uk0 = zk ∀k ∈ K, (13)
∑
k∈K
uk0 =∑
k∈K
zk. (14)
The next constraint states that the PSTN gateway is always present.
z0 = 1. (15)
9
The last five sets of constraints provide the domains for the variables.
s`k ∈ {0, 1} ∀` ∈ L, k ∈ K, (16)
ujk ∈ N ∀j ∈ K, k ∈ K0 \ {j}, (17)
xm` ∈ N ∀m ∈ M, ` ∈ L, (18)
y` ∈ {0, 1} ∀` ∈ L, (19)
zk ∈ {0, 1} ∀k ∈ K0. (20)
2.6 Model Properties
In this section we prove that our problem is NP-hard by showing that it includes the Steiner
tree problem as a special case. Recall that in the Steiner tree problem, which is known to be
NP-hard, one is given a graph G = (V, E) with edge costs for each edge (i, j) ∈ E and the
problem is to find a minimum cost tree that connects a given subset of the nodes U ⊂ V .
The tree may include any of the Steiner nodes V \ U , but is not required to do so.
Proposition 1 The problem (4)–(20) is NP-hard.
Proof Consider the set of instances of our CDMA problem where the input is restricted to
cases where L = M , Cm = {m}, am = 0, bm = 0, dm = 1 ∀m ∈ M , SIRmin < |M |, α > |M |,
and r >∑
m∈M
∑
k∈M cmk +∑
j∈K
∑
k∈K0\{j}hjk. Restrict the input further to cases where
L ⊂ K, and c`k = 0 if ` = k and c`k > r if ` 6= k. Observe that for these problems each
unit of demand is economically attractive to serve since the revenue per channel equivalent
is larger than the cost of building a tower to provide the service and connecting that tower
to the backbone network. Therefore, an optimal solution will serve all of the demand and
10
profit is maximized by finding a minimum cost backbone. The backbone cost is minimized by
connecting tower 1 to MTSO 1, tower 2 to MTSO 2, and so forth, and connecting the PSTN
gateway and MTSOs 1, 2, . . . , M to each other via a minimum cost tree network which may
possibly include some of the other MTSOs. That is, these problems correspond to the set
of all Steiner tree problem instances; G is the graph induced by the MTSOs and the PSTN
gateway, U = {0, 1, 2, . . . , M}, and the cost of edge (i, j) = hij.
In terms of the formulation (4)–(20), xmm = ym = smm = 1 is optimal. Constraints
(5)–(7) and (10) are trivially satisfied, while (18) and (19) are redundant. The problem
reduces to
min∑
m∈M
∑
k∈K
cmksmk
∑
j∈K
∑
k∈K0\{j}
hjkwjk (21)
subject to
∑
k∈K
smk = 1 ∀m ∈ M, (22)
smk ≤ zk ∀m ∈ M, ∀k ∈ K, (23)
ujk ≤ |K|wjk ∀j ∈ K, k ∈ K0 \ {j}, (24)
ujk ≤ |K|zk ∀j ∈ K, k ∈ K0 \ {j}, (25)
∑
j∈K\{k}
(ukj − ujk) + uk0 = zk ∀k ∈ K, (26)
∑
k∈K
uk0 =∑
k∈K
zk, (27)
z0 = 1, (28)
11
smk ∈ {0, 1} ∀m ∈ M, k ∈ K, (29)
ujk ∈ N ∀j ∈ K, k ∈ K0 \ {j}, (30)
zk ∈ {0, 1} ∀k ∈ K0. (31)
This is the flow formulation of the Steiner tree problem where the tower locations and the
PSTN gateway location represent the customer locations and the candidate MTSO locations
represent the Steiner nodes. The Steiner tree problem is known to be NP-hard.
Kalvenes et al. (2003) showed that in the CDMA network design problem, customers
are always assigned to the nearest tower that is constructed so as to minimize overall sys-
tem interference levels. That is, the following set of valid constraints can be added to the
formulation:
xm` ≤ dm(1 − yj) ∀m ∈ M, `, j ∈ Cm such that gm` < gmj. (32)
In order to improve computational performance, we add a set of valid inequalities to
speed up the pruning of the branch-and-bound tree in CPLEX. Constraint (7) limits the
total received signal power at tower `, regardless of the signal source. A subset of the total
received power comes from customers assigned to tower ` for service, i.e., those customer
locations m for which xm` ≥ 1. Thus, if (7) is satisfied, so is the following set of constraints:
∑
m∈P`
xm` ≤ 1 +1
SIRmin
∀` ∈ L. (33)
12
Also note that, in the formulation (4)–(20), the variable s is integer. However, con-
straints (8)–(10) together with the objective function ensure that s is either 0 or 1 even
if the integrality restriction is relaxed. In our computational procedure, we therefore use
0 ≤ s`k ≤ 1 ∀` ∈ L, k ∈ K instead of (16).
3 Empirical Analysis
Our model is implemented in software using the AMPL modeling language (Fourer et al.
2003) with a direct link to the solver in CPLEX (http://www.cplex.com). All test runs are
made on a Compaq AlphaServer DS20E with dual EV 6.7(21264A) 667 MHz processors and
4,096 MB of RAM. Upper and lower bound models are applied to obtain provably near-
optimal solutions for realistic-sized problem instances. The computational times increase
substantially as the number of candidate towers increases from 40 to 160. Therefore, we also
implemented a heuristic solution procedure to solve the largest problems instances.
3.1 Solution Procedure with Error Guarantee
Our solution procedure generates a feasible solution and an upper bound to demonstrate the
quality of the feasible solution. The upper bound procedure solves to optimality the problem
(4)–(33) with the integrality constraint on variables x, y and s relaxed. In the lower bound
procedure, integrality is imposed, but an optimality gap of 5% is permitted.
We created two series of test problems for the empirical evaluation of our proposed
solution method. Both series of test problems were based on the parameters listed in Table
I. While these data do not represent any service provider’s actual system, we have conferred
13
with local service provider engineers to confirm the validity of the parameter value ranges.
Table I about here.
In the first series of test problems, customer demand points and candidate tower loca-
tions were drawn from a uniform distribution over a 13.5 km by 8.5 km rectangular service
area. The number of demand points was 1,000 and 2,000, respectively, while the number of
candidate tower locations was 40, 80, 120, or 160. Six candidate MTSO locations (includ-
ing the PSTN gateway) were drawn from a uniform distribution over a 1.5 km by 1.0 km
rectangular area centered on the 13.5 km by 8.5 km service area. Each demand point had
demand drawn from a uniform distribution of integers between 1 and 10 channel equivalents.
With a mean of 5.5 units of demand in each customer location, the mean demand over the
entire service area was 5,500 and 11,000, respectively. The attenuation factors gm` were then
calculated based on Hata’s path loss model (Hata 1980). A tower location ` that was close
enough to provide service to customer point m (given by the requirement that gm` > 10−15)
would be included in the set Cm. Depending on the number of towers in the service area, the
average size of the sets Cm varied between 2.0 and 8.4. The test problem data are listed in
Table II and problem instance R500 is displayed in Figure 1.
Table II about here.
Figure 1 about here.
The computational results for the forty test problems with randomly distributed cus-
tomer locations are displayed in Table III. The table shows that our solution procedure can
14
find very high quality solutions for realistic-size problems with reasonable compuational ef-
fort. The solution times varied from less than thirty seconds for the smaller problem instances
(R110–R150) to less than sixty minutes for the larger problem instances (R460–R500). Thus,
when we increased the number of customer locations from 1,000 to 2,000 and the number of
candidate tower locations from 40 to 160 (implying a larger number of possible tower selec-
tions for each customer location), the computational effort increased by less than two orders
of magnitude. In the smaller problem instances, one MTSO was selected, while two MTSOs
were in the solution for most of the larger problem instances.
Table III about here.
The upper bound problem was solved to optimality, while the best feasible procedure
was terminated when a solution was found that was less than 5% less than the upper bound
generated by the branch-and-bound tree in CPLEX. Comparing the upper bound solution to
the best feasible solution, we observe that the optimality gap did not increase significantly
as the problem size increased. For nine of the ten largest problem instances (R360–R400 and
R460–R500), the upper bound procedure could not find a solution within 2 hours of CPU-
time. In these cases, we reported the error tolerance (mipgap) of the best feasible solution
procedure (which was 5%). Figure 2 illustrates the solution for test problem R500.
Figure 2 about here.
Next, we solved seven problem instances with data from the North Dallas service area.
We created sample problems with demand points concentrated along the major thoroughfares.
In addition, we created three hot spots of demand in the downtown district, the Galleria area
15
and the DFW airport. Residual demand was drawn from a uniform distribution over the
service area. In each customer location, demand was drawn from a uniform distribution
with values between one and ten simultaneous users. In these problem instances, there are
six candidate MTSO locations, 120 candidate tower locations, and 2,000 customer locations
with the number of simultaneous calls in each location distributed uniformly between one
and ten. Problem ND700 is illustrated in Figure 3.
Figure 3 about here.
The solution to these seven problems are presented in Table IV. We note that the
quality of the solutions as well as the computational times are comparable to those for the
random problem instances in Table III.
Table IV about here.
The solution to test problem ND700 is illustrated in Figure 4. Examining this figure, a
network engineer may find that coverage using the 82 selected towers is insufficient in certain
areas. To remedy this problem, the network engineer can add candidate tower locations
and solve the problem again using the current solution as a starting point. In this example,
we added six towers to the current solution and re-solved the customer allocation problem
with these 88 towers fixed. The CPU time for the modified problem was 1 second and the
resulting solution is displayed in Figure 5. In the modified solution, the coverage increased
from 84.7% to 89.8% and the net revenue increased from $31.93 million to $33.73 million
(or 5.6%). It is possible that this solution is not optimal given the full set of 126 candidate
tower locations. However, a network engineer can use our solution method in an interactive
16
fashion and, when satisfied with the options for candidate tower locations, can solve the
entire problem to optimality. A network engineer can also consciously choose to add towers
in an area where it is not profitable in anticipation of future expansion needs. Thus, our tool
provides considerable flexibility to the network engineer.
Figure 4 about here.
Figure 5 about here.
3.2 Heuristic Procedures
Based on our experience with the computational procedure presented in the section above, we
observed that solution times increase substantially as the average number of towers that can
service a customer area increases. This observation lead us to design two heuristic procedures
that capitalize on limiting the number of towers to which a customer area can be assigned.
The first heuristic solves the problem (4)–(20) with the valid constraints (32) and (33), but
with Cm limited to the nearest tower in the set |L|. The modified test problem data are
displayed in Table V. Since some customer areas are too far from any tower to receive
service, the average number of towers per demand area is slightly below one.
Table V about here.
Table VI gives the computational results for Heuristic 1 compared to the feasible so-
lution procedure presented in the previous section. We observe that Heuristic 1 performs
well on the smaller problem instances, but that the optimality gap increases substantially
for larger problem instances. The reason is that Heuristic 1 will add too many towers to
17
the solution in order to service the customers. While it is better to service these customers
from a larger number of towers than not serving them at all, using such a large number of
towers is inefficient. It is interesting to note, though, that Heuristic 1 performs better on
problem instances with high demand density per tower (i.e., the optimality gap is smaller
for the problem instances with 2,000 customer locations than for those with 1,000 customer
locations with the same number of candidate tower locations). This result stems from the
fact that in the high-density demand problem instances, a higher percentage of the candidate
towers will be constructed in the optimal solution, resulting in a smaller difference in solution
between optimum and the solution obtained with Heuristic 1. Over all, we conclude that
Heuristic 1 is too restrictive in the solution space considered to be of any significant practical
use.
Table VI about here.
In the second heuristic, we restrict the set of permissible tower assignments to at most
two for each customer area. Table VII displays the modified test data for Heuristic 2. Again,
some of the customer service areas are not within the range of two towers (or not within the
range of any tower) and, thus, the average number of towers considered per demand area is
slightly smaller than two.
Table VII about here.
The computational results for Heuristic 2 compared to the feasible solution procedure
are displayed in Table VIII. Since the upper bound procedure failed to produce a solution
within 2 hours of computational time for problem instances R360–R390 and R460-R500, we
18
used the objective function value obtained with the best feasible solution procedure in the
previous section, divided by 0.95 (1-mipgap) to generate an upper bound. Thus, the gap
reported for Heuristic 2 in Table VIII may be larger than the actual gap. However, the
solutions based on Heuristic 2 are not quite as good as the solution obtained with the best
feasible solution procedure. We note that the computational times are shorter for Heuristic
2 than for the best feasible solution procedure from the previous section, in particular for the
larger problem instances. At the same time, the difference in solution quality is less than 5%
in all problem instances. Thus, although Heuristic 2 does not provide an error guarantee, it
is robust enough to produce good feasible solution within reasonable computational times for
very large problem instances. This is particularly true for the higher demand density prob-
lem instances with 160 towers (R460–R500), for which the best feasible solution procedure
requires the most computational time (35–60 minutes compared to 5–10 minutes for Heuris-
tic 2). We conclude that Heuristic 2 is a viable solution procedure for very large problem
instances with high demand density.
Table VIII about here.
4 Conclusions
In this investigation, the W–CDMA network design problem is modeled as a discrete opti-
mization problem. The model maximizes revenue from customers serviced by the network
net the cost of towers, switching facilities and backbone network connecting the towers and
switching facilities to the public switched telephone network. The resulting integer program
is very large and standard commercial software packages cannot obtain optimal solutions to
19
realistic-sized problem instances. Therefore, we developed a solution method based on a pair
of models, one for the upper bound and one for the lower bound. The solution method was
implemented in software using the AMPL/CPLEX system.
We tested our solution method on 40 large test problems with 1,000 to 2,000 customer
locations with an average of 5.5 customers in each location, while the candidate tower lo-
cations varied between 40 and 160 and the number of candidate switching locations was 5.
We solved all of these test problems to within a guaranteed 5% of optimality using very rea-
sonable computational effort. The largest test problems required up to 60 minutes of CPU
time. In an effort to reduce the computational times for the largest and most difficult prob-
lem instances, we developed and tested two heuristic procedures. One of these procedures
proved efficient for the largest test problems, reducing the computational effort by one order
of magnitude at a penalty of less than 5% of the objective function value.
We also tested our solution method on seven test problems based on the infrastructure
and travel patterns in the North Dallas area. The results for these test problems were on
par with those for the randomly generated test problems. Additionally, we provided an
example of how our tool can be used in an interactive fashion in which a network engineer
can manually modify the solution to expand the number of candidate towers or to make use
of specific parts of the network infrastructure. Modifications to a solution can be evaluated
in seconds with our solution method. Thus, it provides network engineers with significant
flexibility when analyzing a network provisioning plan.
20
References
Amaldi, E., A. Capone, and F. Malucelli, 2001a. Discrete Models and Algorithms for
the Capacitated Location Problems Arising in UMTS Network Planning. In Proceedings of
the 5th International Workshop on Discrete Algorithms and Methods for Mobile Computing
and Communications, pp. 1–8, Rome. ACM.
Amaldi, E., A. Capone, and F. Malucelli, 2001b. Improved Models and Algorithms
for UMTS Radio Planning. In IEEE 54th Vehicular Technology Conference Proceedings,
pp. 920–924. IEEE.
Beasley, J. E., 1989. An SST-Based Algorithm for the Steiner Problem in Graphs. Net-
works 19(1), 1–16.
Fourer, R., D. M. Gay, and B. W. Kernighan, 2003. AMPL: A Modeling Language
for Mathematical Programming . Brooks/Cole—Thomson Learning, Pacific Grove, CA,
2nd edn.
Galota, M., C. Glasser, S. Reith, and H. Vollmer, 2001. A Ploynomial-Time
Approximation Scheme for Base Station Positioning in UMTS Networks. In Proceedings of
the 5th International Workshop on Discrete Algorithms and Methods for Mobile Computing
and Communications, pp. 52–59, Rome. ACM.
Hata, M., 1980. Empirical Formula for Propagation Loss in Land Mobile Radio Service.
IEEE Transactions on Vehicular Technology 29, 317–325.
21
Kalvenes, J., J. Kennington, and E. Olinick, 2003. Base Station Location and Service
Assignment in W–CDMA Networks. Technical Report 02-EMIS-03, School of Engineering,
Southern Methodist University, Dallas, TX, http://engr.smu.edu/˜olinick/cdma.pdf.
Li, J., H. Kameda, and H. Itoh, 1997. Balanced Assignment of Cells in PCS Networks.
In Proceedings of the 1997 ACM Symposium on Applied Computing , pp. 297–301, San
Jose, CA. ACM Press.
Martello, S., and P. Toth, 1990. Knapsack Problems: Algorithms and Computer Im-
plementations. Wiley, Chichester, England.
Mathar, R., and M. Schmeink, 2001. Optimal Base Station Positioning and Channel As-
signment for 3G Mobile Networks by Integer Programming. Annals of Operations Research
107, 225–236.
Merchant, A., and B. Sengupta, 1995. Assignment of Cells to Switches in PCS Net-
works. ACM/IEEE Transactions on Networking 3(5), 521–526.
Pisinger, D., 1999. Core Problems in Knapsack Algorithms. Operations Research 47,
570–575.
22
Parameter Value or Range Description
r $4,282 Annual revenue for each customer channel equiva-lent serviced
a` U [$70, 000, $100, 000] Annualized cost for instaling a base station in lo-cation `
bk U [$300, 000, $375, 000] Annualized cost for installing an MTSO in locationk
f $1.00 Annualized cost per foot of wiringα 225 Maximum number of base stations that can be
connected to an MTSO
Table I: Parameters used in the computational experiments.
23
Problem Number of Number of Total Average
Name Candidate Customer Number of Size of
Towers Locations Customers Cm
R110 40 1,000 5,620 2.1R120 40 1,000 5,637 2.0R130 40 1,000 5,638 2.0R140 40 1,000 5,626 2.1R150 40 1,000 5,625 2.2R160 80 1,000 5,609 4.2R170 80 1,000 5,598 4.2R180 80 1,000 5,608 4.1R190 80 1,000 5,593 4.2R200 80 1,000 5,608 4.2R210 120 1,000 5,665 6.4R220 120 1,000 5,676 6.4R230 120 1,000 5,686 6.3R240 120 1,000 5,669 6.4R250 120 1,000 5,695 6.4R410 160 1,000 5,612 8.4R420 160 1,000 5,630 8.4R430 160 1,000 5,628 8.3R440 160 1,000 5,642 8.3R450 160 1,000 5,650 8.3R260 40 2,000 11,020 2.0R270 40 2,000 11,031 2.1R280 40 2,000 11,031 2.1R290 40 2,000 11,036 2.1R300 40 2,000 11,051 2.1R310 80 2,000 11,038 4.3R320 80 2,000 11,051 4.3R330 80 2,000 11,041 4.3R340 80 2,000 11,033 4.3R350 80 2,000 11,033 4.3R360 120 2,000 11,042 6.4R370 120 2,000 11,030 6.4R380 120 2,000 11,033 6.4R390 120 2,000 11,015 6.4R400 120 2,000 11,010 6.4R460 160 2,000 11,001 8.4R470 160 2,000 11,003 8.4R480 160 2,000 11,001 8.4R490 160 2,000 10,998 8.4R500 160 2,000 10,987 8.4
Table II: Problem data for uniformly distributed subscribers (dm ∼ U [1, 10]).
24
Problem Upper Bound Best Feasible Solution (mipgap=5%)Name MTSOs Towers Customers Profit CPU Time MTSOs Towers Customers Profit CPU Time Optimality
Built Built Serviced ($M) (hh:mm:ss) Built Built Serviced ($M) (hh:mm:ss) Gap
R110 1 35.6 92.6% 18.33 00:00:02 1 37 92.8% 18.22 00:00:20 0.6%R120 1 35.7 92.4% 18.37 00:00:03 1 36 92.2% 18.30 00:00:21 0.4%R130 1 33.4 88.4% 17.63 00:00:02 1 35 88.8% 17.53 00:00:16 0.6%R140 1 34.0 86.4% 17.03 00:00:29 1 35 85.9% 16.80 00:00:25 1.3%R150 1 30.9 85.0% 17.09 00:00:03 1 31 84.8% 16.99 00:00:22 0.6%R160 1 42.0 92.2% 17.55 00:08:43 1 39 87.5% 16.74 00:01:40 4.6%R170 1 42.0 91.4% 17.36 00:05:41 1 43 89.3% 16.75 00:01:50 3.6%R180 1 41.0 92.4% 17.79 00:10:34 1 43 90.4% 17.10 00:01:38 4.0%R190 1 42.7 92.8% 17.61 00:05:40 1 44 91.1% 17.07 00:01:46 3.2%R200 1 43.5 92.1% 17.41 00:10:58 1 43 89.7% 16.87 00:01:50 3.2%R210 1 50.0 94.2% 17.66 00:43:18 1 51 91.5% 16.97 00:08:48 4.1%R220 1 51.1 94.7% 17.58 00:26:28 1 53 92.6% 16.92 00:07:32 3.9%R230 1 50.1 94.9% 17.88 00:24:44 1 48 91.0% 17.15 00:05:05 4.3%R240 1 48.0 94.0% 17.81 00:24:33 1 51 93.0% 17.27 00:06:29 3.1%R250 1 48.1 94.3% 17.91 00:45:44 1 48 92.2% 17.42 00:07:54 2.9%R410 1 53.1 93.1% 16.81 00:57:02 1 53 90.3% 16.21 00:15:07 3.7%R420 1 53.7 93.1% 16.78 01:40:11 1 54 90.2% 16.04 00:30:43 4.6%R430 1 52.3 92.7% 16.85 01:44:26 1 53 90.2% 16.23 00:21:03 3.9%R440 1 54.0 93.2% 16.86 01:00:32 1 55 90.8% 16.25 00:17:25 3.7%R450 2 55.8 94.2% 16.84 01:40:19 1 55 91.4% 16.16 00:15:53 4.2%R260 1 37.0 65.3% 26.72 00:00:14 1 38 65.3% 26.60 00:01:17 0.4%R270 1 37.6 67.6% 27.80 00:00:14 1 39 67.7% 27.70 00:00:50 0.4%R280 1 37.3 69.7% 28.85 00:00:17 1 38 69.7% 28.75 00:00:52 0.4%R290 1 37.4 68.2% 28.25 00:00:17 1 39 68.3% 28.12 00:01:19 0.5%R300 1 38.2 66.8% 27.63 00:00:14 1 40 67.0% 27.53 00:00:59 0.4%R310 1 62.4 87.6% 34.93 00:10:04 1 65 86.8% 34.33 00:03:51 1.8%R320 1 62.0 87.0% 34.79 00:07:44 1 57 83.7% 33.70 00:03:32 3.2%R330 1 64.0 88.3% 35.14 00:09:17 1 68 87.9% 34.50 00:04:12 1.9%R340 1 65.1 89.1% 35.32 00:10:38 1 66 87.7% 34.59 00:04:16 2.1%R350 1 64.5 88.6% 35.15 00:34:46 1 64 87.2% 34.58 00:04:20 1.7%
R360 N/A N/A N/A N/A 02:00:00† 1 75 93.4% 36.42 00:14:52 5.0%R370 N/A N/A N/A N/A 02:00:00† 2 81 94.5% 36.38 00:16:22 5.0%R380 N/A N/A N/A N/A 02:00:00† 2 80 92.2% 35.27 00:17:14 5.0%R390 N/A N/A N/A N/A 02:00:00† 2 78 93.6% 36.15 00:21:53 5.0%R400 2 76.7 94.1% 37.16 00:59:47 2 79 94.1% 36.26 00:15:29 2.5%
R460 N/A N/A N/A N/A 02:00:00† 2 88 93.7% 35.24 00:56:40 5.0%R470 N/A N/A N/A N/A 02:00:00† 1 84 91.4% 34.61 00:35:20 5.0%R480 N/A N/A N/A N/A 02:00:00† 2 88 93.6% 35.25 00:36:12 5.0%R490 N/A N/A N/A N/A 02:00:00† 2 92 94.7% 35.31 00:41:08 5.0%R500 N/A N/A N/A N/A 02:00:00† 2 87 93.1% 34.95 00:49:15 5.0%
† terminated due to 2-hour time limit.
Table III: Empirical results for test problems with uniformly distributed subscribers (dm ∼ U [1, 10]).
25
Problem Upper Bound Best Feasible Solution (mipgap=5%)Name MTSOs Towers Customers Profit CPU Time MTSOs Towers Customers Profit CPU Time Optimality
Built Built Serviced ($M) (hh:mm:ss) Built Built Serviced ($M) (hh:mm:ss) Gap
ND100 2 75.4 82.6% 31.55 00:39:05 2 77 82.0% 31.13 00:02:12 1.4%ND200 2 81.3 85.5% 32.33 00:55:18 3 83 84.8% 31.71 00:03:35 2.0%ND300 3 84.2 87.3% 33.00 01:13:05 2 82 85.2% 32.17 00:03:15 2.6%ND400 2 79.5 86.3% 32.92 00:45:29 2 80 85.6% 32.45 00:03:04 1.5%ND500 2 81.5 87.0% 32.93 00:40:46 2 82 85.0% 32.00 00:03:10 2.9%ND600 2 80.7 86.1% 32.86 00:46:06 2 82 85.9% 32.63 00:06:00 0.7%ND700 2 81.4 85.6% 32.36 00:53:35 2 82 84.7% 31.93 00:03:31 1.4%
Table IV: Empirical results for North Dallas test problems.
26
Problem Number of Number of Total Average
Name Candidate Customer Number of Size of
Towers Locations Customers Cm
R110 40 1,000 5,620 0.94R120 40 1,000 5,637 0.94R130 40 1,000 5,638 0.90R140 40 1,000 5,626 0.88R150 40 1,000 5,625 0.87R160 80 1,000 5,609 0.96R170 80 1,000 5,598 0.96R180 80 1,000 5,608 0.96R190 80 1,000 5,593 0.96R200 80 1,000 5,608 0.96R210 120 1,000 5,665 0.99R220 120 1,000 5,676 0.99R230 120 1,000 5,686 1.00R240 120 1,000 5,669 1.00R250 120 1,000 5,695 1.00R410 160 1,000 5,612 1.00R420 160 1,000 5,630 1.00R430 160 1,000 5,628 1.00R440 160 1,000 5,642 1.00R450 160 1,000 5,650 1.00R260 40 2,000 11,020 0.90R270 40 2,000 11,031 0.90R280 40 2,000 11,031 0.90R290 40 2,000 11,036 0.87R300 40 2,000 11,051 0.81R310 80 2,000 11,038 0.97R320 80 2,000 11,051 0.98R330 80 2,000 11,041 0.98R340 80 2,000 11,033 0.97R350 80 2,000 11,033 0.97R360 120 2,000 11,042 0.98R370 120 2,000 11,030 0.99R380 120 2,000 11,033 0.99R390 120 2,000 11,015 0.99R400 120 2,000 11,010 0.99R460 160 2,000 11,001 1.00R470 160 2,000 11,003 1.00R480 160 2,000 11,001 1.00R490 160 2,000 10,998 1.00R500 160 2,000 10,987 1.00
Table V: Problem data for uniformly distributed subscribers (dm ∼ U [1, 10]) for Heuristic 1with |Cm| ≤ 1. 27
Problem Best Feasible Solution (mipgap=5%) Heuristic 1 (|Cm| ≤ 1)Name MTSOs Towers Customers Profit CPU Time Optimality MTSOs Towers Customers Profit CPU Time Optimality
Built Built Serviced ($M) (hh:mm:ss) Gap Built Built Serviced ($M) (hh:mm:ss) Gap
R110 1 37 92.8% 18.22 00:00:20 0.6% 1 40 93.5% 18.09 00:00:01 1.3%R120 1 36 92.2% 18.30 00:00:21 0.4% 1 39 92.6% 18.08 00:00:01 1.6%R130 1 35 88.8% 17.53 00:00:16 0.6% 1 39 89.5% 17.28 00:00:01 2.0%R140 1 35 85.9% 16.80 00:00:25 1.3% 1 38 86.6% 16.69 00:00:01 2.0%R150 1 31 84.8% 16.99 00:00:22 0.6% 1 40 86.3% 16.48 00:00:01 3.6%R160 1 39 87.5% 16.74 00:01:40 4.6% 1 67 92.4% 15.05 00:00:01 14.2%R170 1 43 89.3% 16.75 00:01:50 3.6% 1 67 91.4% 14.90 00:00:01 14.2%R180 1 43 90.4% 17.10 00:01:38 4.0% 1 70 94.0% 15.21 00:00:01 14.5%R190 1 44 91.1% 17.07 00:01:46 3.2% 1 69 93.2% 15.06 00:00:01 14.4%R200 1 43 89.7% 16.87 00:01:50 3.2% 2 69 92.5% 15.95 00:00:01 8.3%R210 1 51 91.5% 16.97 00:08:48 4.1% 2 94 93.0% 13.03 00:00:01 26.2%R220 1 53 92.6% 16.92 00:07:32 3.9% 2 98 95.0% 13.10 00:00:01 25.5%R230 1 48 91.0% 17.15 00:05:05 4.3% 2 96 94.0% 13.20 00:00:01 26.2%R240 1 51 93.0% 17.27 00:06:29 3.1% 2 98 94.7% 13.02 00:00:01 26.9%R250 1 48 92.2% 17.42 00:07:54 2.9% 2 95 94.5% 13.40 00:00:01 25.2%R410 1 53 90.3% 16.21 00:15:07 3.7% 1 94 83.9% 10.53 00:00:01 37.4%R420 1 54 90.2% 16.04 00:30:43 4.6% 2 102 86.4% 10.42 00:00:01 37.9%R430 1 53 90.2% 16.23 00:21:03 3.9% 2 104 87.4% 10.63 00:00:01 36.9%R440 1 55 90.8% 16.25 00:17:25 3.7% 2 106 89.4% 10.79 00:00:01 36.0%R450 1 55 91.4% 16.16 00:15:53 4.2% 2 104 88.0% 10.62 00:00:01 36.9%R260 1 38 65.3% 26.60 00:01:17 0.4% 1 40 65.3% 26.38 00:00:03 1.3%R270 1 39 67.7% 27.70 00:00:50 0.4% 1 40 67.0% 27.29 00:00:03 1.8%R280 1 38 69.7% 28.75 00:00:52 0.4% 1 40 69.7% 28.55 00:00:05 1.0%R290 1 39 68.3% 28.12 00:01:19 0.5% 1 40 68.0% 27.91 00:00:03 1.2%R300 1 40 67.0% 27.53 00:00:59 0.4% 1 40 66.6% 27.35 00:00:02 1.0%R310 1 65 86.8% 34.33 00:03:51 1.8% 1 79 89.8% 33.90 00:00:02 2.9%R320 1 57 83.7% 33.70 00:03:32 3.2% 2 77 87.7% 33.62 00:00:02 3.4%R330 1 68 87.9% 34.50 00:04:12 1.9% 2 80 89.6% 34.15 00:00:02 2.8%R340 1 66 87.7% 34.59 00:04:16 2.1% 2 80 90.3% 34.47 00:00:02 2.4%R350 1 64 87.2% 34.58 00:04:20 1.7% 2 78 89.8% 34.27 00:00:03 2.5%R360 1 75 93.4% 36.42 00:14:52 5.0% 2 113 96.5% 34.39 00:00:01 10.3%R370 2 81 94.5% 36.38 00:16:22 5.0% 2 113 96.6% 34.34 00:00:01 10.3%R380 2 80 92.2% 35.27 00:17:14 5.0% 2 117 96.7% 33.96 00:00:01 8.5%R390 2 78 93.6% 36.15 00:21:53 5.0% 2 117 96.5% 33.86 00:00:01 11.0%R400 2 79 94.1% 36.26 00:15:29 2.5% 2 115 96.8% 34.11 00:00:01 8.2%R460 2 88 93.7% 35.24 00:56:40 5.0% 2 141 96.4% 31.51 00:00:01 15.1%R470 1 84 91.4% 34.61 00:35:20 5.0% 2 140 95.7% 31.26 00:00:01 14.2%R480 2 88 93.6% 35.25 00:36:12 5.0% 2 140 96.0% 31.39 00:00:01 15.4%R490 2 92 94.7% 35.31 00:41:08 5.0% 2 138 95.7% 31.46 00:00:01 15.4%R500 2 87 93.1% 34.95 00:49:15 5.0% 2 142 96.4% 31.20 00:00:01 15.2%
Table VI: Empirical results for Heuristic 1 applied to test problems with uniformly distributed subscribers (dm ∼ U [1, 10]).
28
Problem Number of Number of Total Average
Name Candidate Customer Number of Size of
Towers Locations Customers Cm
R110 40 1,000 5,620 1.62R120 40 1,000 5,637 1.61R130 40 1,000 5,638 1.51R140 40 1,000 5,626 1.48R150 40 1,000 5,625 1.50R160 80 1,000 5,609 1.81R170 80 1,000 5,598 1.81R180 80 1,000 5,608 1.83R190 80 1,000 5,593 1.81R200 80 1,000 5,608 1.85R210 120 1,000 5,665 1.95R220 120 1,000 5,676 1.95R230 120 1,000 5,686 1.96R240 120 1,000 5,669 1.95R250 120 1,000 5,695 2.00R410 160 1,000 5,612 1.97R420 160 1,000 5,630 1.96R430 160 1,000 5,628 1.97R440 160 1,000 5,642 1.97R450 160 1,000 5,650 1.97R260 40 2,000 11,020 1.49R270 40 2,000 11,031 1.52R280 40 2,000 11,031 1.51R290 40 2,000 11,036 1.48R300 40 2,000 11,051 1.42R310 80 2,000 11,038 1.84R320 80 2,000 11,051 1.85R330 80 2,000 11,041 1.87R340 80 2,000 11,033 1.88R350 80 2,000 11,033 1.88R360 120 2,000 11,042 1.95R370 120 2,000 11,030 1.95R380 120 2,000 11,033 1.95R390 120 2,000 11,015 1.95R400 120 2,000 11,010 1.94R460 160 2,000 11,001 1.98R470 160 2,000 11,003 1.97R480 160 2,000 11,001 1.97R490 160 2,000 10,998 1.96R500 160 2,000 10,987 1.96
Table VII: Problem data for uniformly distributed subscribers (dm ∼ U [1, 10]) for Heuristic2 with |Cm| ≤ 2. 29
Problem Best Feasible Solution (mipgap=5%) Heuristic 2 (|Cm| ≤ 2)Name MTSOs Towers Customers Profit CPU Time Optimality MTSOs Towers Customers Profit CPU Time Optimality
Built Built Serviced ($M) (hh:mm:ss) Gap Built Built Serviced ($M) (hh:mm:ss) Gap
R110 1 37 92.8% 18.22 00:00:20 0.6% 1 37 92.8% 18.22 00:00:14 0.6%R120 1 36 92.2% 18.30 00:00:21 0.4% 1 36 92.2% 18.30 00:00:15 0.4%R130 1 35 88.8% 17.53 00:00:16 0.6% 1 35 88.8% 17.53 00:00:10 0.6%R140 1 35 85.9% 16.80 00:00:25 1.3% 1 35 86.3% 16.90 00:00:14 0.7%R150 1 31 84.8% 16.99 00:00:22 0.6% 1 34 85.6% 16.90 00:00:13 0.6%R160 1 39 87.5% 16.74 00:01:40 4.6% 1 47 90.4% 16.53 00:00:33 3.6%R170 1 43 89.3% 16.75 00:01:50 3.6% 1 52 91.7% 16.44 00:01:00 3.2%R180 1 43 90.4% 17.10 00:01:38 4.0% 1 49 91.7% 16.87 00:00:46 3.1%R190 1 44 91.1% 17.07 00:01:46 3.2% 1 49 90.9% 16.52 00:00:33 4.0%R200 1 43 89.7% 16.87 00:01:50 3.2% 1 53 92.0% 16.47 00:01:00 3.3%R210 1 51 91.5% 16.97 00:08:48 4.1% 1 67 93.9% 15.88 00:01:14 4.1%R220 1 53 92.6% 16.92 00:07:32 3.9% 2 66 94.0% 16.06 00:01:34 3.0%R230 1 48 91.0% 17.15 00:05:05 4.3% 1 66 94.6% 16.11 00:01:42 3.8%R240 1 51 93.0% 17.27 00:06:29 3.1% 1 63 92.7% 16.08 00:01:08 3.6%R250 1 48 92.2% 17.42 00:07:54 2.9% 1 65 93.4% 15.99 00:01:15 4.6%R410 1 53 90.3% 16.21 00:15:07 3.7% 1 76 92.1% 14.26 00:00:54 14.3%R420 1 54 90.2% 16.04 00:30:43 4.6% 1 77 92.3% 14.11 00:01:30 15.9%R430 1 53 90.2% 16.23 00:21:03 3.9% 2 77 91.8% 14.20 00:00:51 15.7%R440 1 55 90.8% 16.25 00:17:25 3.7% 1 75 91.7% 14.32 00:00:49 15.1%R450 1 55 91.4% 16.16 00:15:53 4.2% 1 77 93.6% 14.69 00:01:11 12.8%R260 1 38 65.3% 26.60 00:01:17 0.4% 1 38 65.3% 26.60 00:00:45 0.4%R270 1 39 67.7% 27.70 00:00:50 0.4% 1 39 67.7% 27.71 00:00:35 0.4%R280 1 38 69.7% 28.75 00:00:52 0.4% 1 38 69.7% 28.75 00:00:36 0.4%R290 1 39 68.3% 28.12 00:01:19 0.5% 1 39 68.2% 28.06 00:00:45 0.7%R300 1 40 67.0% 27.53 00:00:59 0.4% 1 40 67.0% 27.53 00:00:36 0.4%R310 1 65 86.8% 34.33 00:03:51 1.8% 1 65 86.5% 34.21 00:01:52 1.8%R320 1 57 83.7% 33.70 00:03:32 3.2% 1 65 86.4% 34.17 00:01:21 1.4%R330 1 68 87.9% 34.50 00:04:12 1.9% 1 67 87.5% 34.46 00:01:35 1.8%R340 1 66 87.7% 34.59 00:04:16 2.1% 1 68 88.2% 34.61 00:01:24 1.9%R350 1 64 87.2% 34.58 00:04:20 1.7% 1 64 86.6% 34.30 00:01:58 2.3%R360 1 75 93.4% 36.42 00:14:52 5.0% 1 85 94.3% 35.98 00:03:36 6.1%R370 2 81 94.5% 36.38 00:16:22 5.0% 2 87 94.3% 35.73 00:05:14 6.7%R380 2 80 92.2% 35.27 00:17:14 5.0% 2 83 93.6% 35.65 00:04:16 4.0%R390 2 78 93.6% 36.15 00:21:53 5.0% 2 83 93.9% 35.69 00:04:15 6.2%R400 2 79 94.1% 36.26 00:15:29 2.5% 2 84 93.6% 35.63 00:05:03 4.1%R460 2 88 93.7% 35.24 00:56:40 5.0% 2 100 93.3% 33.99 00:06:23 8.4%R470 1 84 91.4% 34.61 00:35:20 5.0% 2 103 94.8% 34.35 00:09:43 5.7%R480 2 88 93.6% 35.25 00:36:12 5.0% 2 101 92.9% 33.69 00:07:02 9.2%R490 2 92 94.7% 35.31 00:41:08 5.0% 2 104 93.7% 33.75 00:05:48 9.2%R500 2 87 93.1% 34.95 00:49:15 5.0% 2 101 93.0% 33.52 00:07:51 8.9%
Table VIII: Empirical results for Heuristic 2 applied to test problems with uniformly distributed subscribers (dm ∼ U [1, 10]).
30
0
1
2
3
4
5
1
2
3
4
5
67
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
9192
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135 136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
Figure 1: Graphical representation of problem R500 with demand locations uniformly dis-tributed over the service area (circles are subscriber locations, triangles are candidate towerlocations and squares are candidate MTSO locations).
31
0
1
3
2
4
5
3
6
9
10
12
13
14
15
1618
19
21
22
24
27
28
29
30
32
33
37
38
39
40
41
42
45
4648
51
53
57
58
60
61
65
66
70
71
72
75
77
80
81
8589
90
95
97
99
100
101
102
105
106
107
109
110
111
112
113
114
115
116
117
119
120
122
127
129
130
131
132
134
136
139
142
143
146
147
148
150
151
152
154
156
160
1
2
4
5
7
8
11
17
20
23
25
26
31
34
35
36
43
44
47
49
50
52
54
55
56
59
62
63
64
67
68
69
73
74
76
78
79
82
83
84
86
87
88
9192
93
94
96
98
103
104
108
118
121
123
124
125
126
128
133
135
137
138
140
141
144
145
149
153
155
157
158
159
Figure 2: Graphical representation of the solution to problem R500.
32
0
1
2
3
4
5
6
1
2
3
45
6 7 8 9 10 11 12 13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41 42
43
44
45
46
47
48
49
50
51
52
53
54
55
5657
58
59
6061
62
63
64
65
6667
68
69
70
71
72
73
74
75
76
77
78
79
8081
82
8384
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106 107
108
109
110
111
112
113
114
115
116
117
118
119
120
Figure 3: Graphical representation of problem ND700 with demand locations concentratedalong four major thoroughfares and in three hotspots in the North Dallas area.
33
0
3
6
1
2
4
5
3
45
6 7 8 9 11 12 13
14
15
16
17
18
20
22
23
24
25
27
28
30
31
32
33
34
35
36
38
39
42
44
45
46
47
4850
51
53
54
55
5657
59
6061
6367
68
70
71
74
75
77
78
79
81
82
83
87
88
89
90
91
93
94
99
101
102
103
105
107
108
112
113
114
116
117
118
119
120
1
210
19
21
26
29
37
40
41
43
49
52
58
62
64
65
66
69
72
73
7680
84
85
86
92
95
96
97
98
100
104
106
109
110
111
115
Figure 4: Graphical representation of the solution to problem ND700.
34
0
3
6
1
2
4
5
3
45
6 7 8 9 11 12 13
14
15
16
17
18
20
22
23
24
25
27
28
30
31
32
33
34
35
36
38
39
42
44
45
46
47
4850
51
53
54
55
5657
59
6061
6367
68
70
71
74
75
77
78
79
81
82
83
87
88
89
90
91
93
94
99
101
102
103
105
107
108
112
113
114
116
117
118
119
120
121
122
123
124
125
126
1
210
19
21
26
29
37
40
41
43
49
52
58
62
64
65
66
69
72
73
7680
84
85
86
92
95
96
97
98
100
104
106
109
110
111
115
Figure 5: Graphical representation of the modified solution to problem ND700.
35