Professional Documents
Culture Documents
10 1287@trsc 2014 0569
10 1287@trsc 2014 0569
Transportation Science
Publication details, including instructions for authors and subscription information:
http://pubsonline.informs.org
This article may be used only for the purposes of research, teaching, and/or private study. Commercial use
or systematic downloading (by robots or other automatic processes) is prohibited without explicit Publisher
approval, unless otherwise noted. For more information, contact permissions@informs.org.
The Publisher does not warrant or guarantee the article’s accuracy, completeness, merchantability, fitness
for a particular purpose, or non-infringement. Descriptions of, or references to, products or publications, or
inclusion of an advertisement in this article, neither constitutes nor implies a guarantee, endorsement, or
support of claims made of that product, publication, or service.
INFORMS is the largest professional society in the world for professionals in the fields of operations research, management
science, and analytics.
For more information on INFORMS, its publications, membership, or meetings visit http://www.informs.org
Vol. 49, No. 2, May 2015, pp. 239–253
ISSN 0041-1655 (print) ISSN 1526-5447 (online) http://dx.doi.org/10.1287/trsc.2014.0569
© 2015 INFORMS
Cristián E. Cortés
Civil Engineering Department, Universidad de Chile, 8370449 Santiago, Chile, ccortes@ing.uchile.cl
Alfredo Núñez
Section of Road and Railway Engineering, Delft University of Technology, 2628 CN Delft, The Netherlands,
a.a.nunezvicencio@tudelft.nl
T his paper presents a methodology based on generic evolutionary algorithms to solve a dynamic pickup and
delivery problem formulated under a hybrid predictive control approach. The solution scheme is designed
to support the dispatcher of a dial-a-ride service, where quick and efficient real-time solutions are needed.
The scheme considers different configurations of particle swarm optimization and genetic algorithms within a
proposed ad-hoc methodology to solve in real time the nonlinear mixed-integer optimization problem related
with the hybrid predictive control approach. These consist of different techniques to handle the operational
constraints (penalization, Baldwinian, and Lamarckian repair) and encodings (continuous and integer). For
parameter tuning, a new approach based on multiobjective optimization is proposed and used to select and
study some of the evolutionary algorithms. The multiobjective feature arises when deciding the parameters with
the best trade-off between performance and computational effort. Simulation results are presented to compare
the different schemes proposed and to advise conditions for the application of the method in real instances.
Keywords: predictive control; dynamic pickup and delivery problem; evolutionary algorithms
History: Received: May 2010; revision received: October 2013; accepted: June 2014. Published online in Articles
in Advance March 10, 2015.
1. Introduction and Background to make routing decisions for a fixed fleet of vehi-
The dynamic pickup and delivery problem (DPDP) cles with limited capacity, operating in a real-time ser-
can be formulated as a set of transportation requests vice where the demand (represented by passengers) is
(identified by pickup and delivery locations) that are unknown in advance. Eksioglu, Volkan, and Reisman
served by a fleet of vehicles initially located at sev- (2009) and Berbeglia, Cordeau, and Laporte (2010)
eral depots. The dynamic dimension appears when a presented comprehensive reviews of DPDP, dial-a-
subset of the requests is not known in advance; there- ride applications, and solution methods.
fore, such requests have to be scheduled for service in Xiang, Chu, and Chen (2008) studied a DRS by
real time, at the instant they call. The DPDP is now of considering a complex set of constraints on a time-
great interest for practitioners and researchers because dependent network. With regard to real applications,
of the development and implementation of efficient Madsen, Raven, and Rygaard (1995) adapted the inser-
online optimization tools; this fact is crucial for the tion heuristics by Jaw et al. (1986) to solve a real-
emerging improvement in the quality of the formula- life problem for moving elderly and handicapped
tion solutions. The DPDP has been intensely studied people in Copenhagen, and Dial (1995) proposed
over the last 20 years (Psaraftis 1980, 1988; Bertsimas an approach to the many-to-few dial-a-ride transit
and Van Ryzin 1991; Bertsimas and Van Ryzin 1993a, operation ADART (Autonomous Dial-a-Ride Transit),
b; Kleywegt and Papastavrou 1998; Gendreau et al. which is currently implemented in Corpus Christi,
1999; Swihart and Papastavrou 1999; Larsen 2000; Texas. Gendreau et al. (1999) modified the tabu-search
Thomas and White 2004). The output of such a prob- heuristics to solve the dynamic vehicle routing prob-
lem should be a set of routes for all of the vehicles, lem (DVRP) with soft time windows in an effort to find
which change dynamically over time. The solution of a solution method that will handle different DVRPs.
a DPDP can be linked with the control of a typical More sophisticated tabu-search methods were recently
dial-a-ride system (DRS), where the dispatcher has developed, such as granular tabu search (Toth and
239
Muñoz-Carpintero et al.: Methodology Based on Evolutionary Algorithms to Solve DPDP
240 Transportation Science 49(2), pp. 239–253, © 2015 INFORMS
Vigo 2003) and adaptive memory based on tabu search ant colonies, have also been applied to solve DVRP
(Tarantilis 2005). Xiang, Chu, and Chen (2008) devel- (Montemanni et al. 2005; Dréo et al. 2006).
oped a heuristic local search strategy that uses a sec- In the current work, the main objective is to present
ondary objective function to drive the search out of an efficient and systematic ad-hoc methodology to
local optima. In the context of DPDP, Mitrovic-Minic, solve the HPC formulation of the DPDP based on
Krishnamurti, and Laporte (2004) introduced the con- generic evolutionary algorithms. The solution scheme
cept of double-horizon based heuristics for solving the will help dispatchers of the dial-a-ride service, where
Downloaded from informs.org by [128.122.253.212] on 23 May 2015, at 06:02 . For personal use only, all rights reserved.
DPDP with time windows, showing that the method efficient real-time solutions are needed quickly to
can yield gains in route costs when compared with make the system work. Specific implementations of
classical (single) rolling horizon methods, but the GA and PSO will result in several variants of the
improvement tends to decrease as instances become generic evolutionary algorithms. In addition, a mul-
larger. Mitrovic-Minic and Laporte (2004) presented tiobjective approach for tuning the parameters of the
four waiting strategies for vehicles (drive-first, wait- proposed evolutionary algorithms is presented. The
first, dynamic waiting, and the advanced dynamic multiobjective feature arises for deciding the parame-
waiting). They concluded that in terms of total route ters with the best trade-off between performance and
length, the proposed strategies outperform the com- computation time required for real applications. A
monly used drive-first waiting strategy, making the detailed analysis of the accuracy and computational
advanced dynamic waiting strategy the most efficient. time is conducted, which can be extended to other
The dial-a-ride system can be modeled for designing complex nonlinear engineering problems containing
a hybrid predictive control (HPC) scheme, consider- mixed integers and continuous variables.
ing that potential rerouting of vehicles could affect the In §2, a summary of the problem statement pro-
current decisions, by analyzing the extra cost of insert- posed by Cortés et al. (2009) is presented, which is the
ing real-time service requests into predefined vehicle starting point for the methodology based on generic
routes while the vehicles are in service. In previous evolutionary algorithms proposed in this work. In §3
works of our group, a formulation of the DPDP in we present the ad-hoc methodology, the PSO and GA
an HPC by specifying the state space variables and configurations used within, and the parameter tun-
models was presented (Sáez, Cortés, and Núñez 2008; ing based on multiobjective optimization. Next, in §4
Cortés, Sáez, and Núñez 2008; Cortés et al. 2009). In a detailed computational analysis and comparison of
those works, two solution algorithms using genetic the configurations of PSO and GA are carried out
algorithms (GA) and particle swarm optimization based on simulations. Finally, conclusions, remarks,
(PSO) were developed to solve real-time instances. To and further research are presented in §5.
the best of our knowledge, no other hybrid predictive
control approaches for solving DPDP have been pro-
posed in the literature using PSO and GA that can per- 2. Problem Statement
form real-time control on a dial-a-ride type of system.
In the literature, most of the applications using 2.1. General Description
such methods (namely, PSO and GA) solve static In the context of control theory, the notion of hybrid
cases, or vehicle routing problems (VRP) that nei- systems arises when the problem conditions are char-
ther include explicitly the dynamic behavior of the acterized by both continuous and discrete/integer
system, nor a reasonable set of future realizations variables. In the last two decades, hybrid systems
of the stochastic demand. Specifically, GAs have have been studied more intensely by researchers from
been applied for various VRPs, considering different several study areas, such as computer science and
chromosome representations and genetic operators automatic control (see for example Bemporad and
depending on the particular problem: Skrlec, Filipec, Morari 1999; Hegyi, De Schutter, and Hellendoorn
and Krajcar (1997), for the single vehicle capacitated 2005; Karer et al. 2007a, b; Núñez et al. 2009). Specifi-
VRP; Haghani and Jung (2005), for the multivehi- cally, hybrid systems can be expressed as a nonlinear
cle DVRP with time-dependent travel time and soft state space model given by
time windows. Zhu et al. (2006) proposed an adapted
PSO algorithm to solve a static VRP with time win- x4k + 15 = f 4x4k51 u4k551
(1)
dows. Jih and Hsu (1999) and Osman, Abo-Sinna, y4k5 = g4x4k51 u4k551
and Mousa (2005) presented a successful comparison
of the GA against dynamic programming in terms where x4k5 are the continuous and/or discrete (inte-
of computation time; the former solved the DVRP ger) state space variables, u4k5 are the continuous
with time windows and capacity constraints and the and/or discrete input or manipulated variables, y4k5
latter solved a multiobjective VRP. Ant colony meth- define the continuous and/or discrete system outputs
ods, a metaheuristic inspired by the behavior of real and f 4·1 ·5, g4·1 ·5 are piecewise nonlinear functions. In
Muñoz-Carpintero et al.: Methodology Based on Evolutionary Algorithms to Solve DPDP
Transportation Science 49(2), pp. 239–253, © 2015 INFORMS 241
general, a hybrid predictive control controller mini- where rji 4k5 is a binary variable defined as follows:
mizes the following generic objective function: (
1 if stop i belonging to Sj 4k5 is a pick-up
min J 4u4k51 0 0 0 1 u4k + Nu − 151 x̂4k + 151 0 0 0 1 rji 4k5 =
u4k510001u4k+Nu −15 0 if stop i belonging to Sj 4k5 is a delivery0
x̂4k + N 51 ŷ4k + 151 0 0 0 1 ŷ4k + N 551 (2)
The first and second columns represent a pair iden-
where J is an objective function, k is the current time, tifying if stop i is either a pickup [1 0] or a deliv-
Downloaded from informs.org by [128.122.253.212] on 23 May 2015, at 06:02 . For personal use only, all rights reserved.
N is the prediction horizon, Nu is the control hori- ery [0 1], respectively. The third column of the Sj 4k5
zon, x̂4k + t5, ŷ4k + t5 are the expected state space vec- matrix represents the external travel time function,
tor and the expected system output at instant k + t, where âji 4k5 is the expected total travel time between
respectively, and 6u4k5T 1 0 0 0 1 u4k + Nu − 15T 7T repre- points i − 1 and i plus the transfer operation delay
sents the control sequence, which corresponds to the at node i. For simulation purposes, we assume that
set of optimization variables. Once expression (2) is the position of the vehicles can be measured or esti-
optimized, only the first element of the control vector mated at any moment. The last column labelij keeps
u4k5 is used to update the system conditions, based the passenger identifier, which is needed to check the
on the receding horizon methodology. feasibility of the sequence in terms of precedence (the
Conceptually, the HPC framework to model the pickup must occur before the delivery of the same
DPDP incorporates stochasticity into the routing dis- client).
patch rules by considering the impact of future reas- When the dispatcher makes a decision, first the pas-
signments on the performance of already-scheduled sengers are assigned to a certain vehicle, and then
customers (Cortés et al. 2009). The stochastic predic- they are inserted within the sequence of task that
tion allows the dispatcher to incorporate a more real- the vehicle follows. Figure 1 shows an example of
istic and robust measure of effective travel (waiting) a sequence for a vehicle j. Users labeled as “1,”
time experienced by the users into the decision objec- “2,” and “3” are assigned to vehicle j. The sequence
tive function. assigned considers to pickup user “label1j 4k5 = 1”
We consider a fleet of F vehicles, which are dynam- (coordinate 1+ ), then to pickup user “label2j 4k5 = 3”
ically routed over an influence area A. The demand (coordinate 3+ ), then to delivery user “label3j 4k5 = 1”
for service is unknown and is revealed in real time. (coordinate 1− ) and so on.
Quick routing and scheduling decisions are required
Vehicles will travel according to the predefined
to handle the demand with the available vehicles. At
sequence vector S4k − 15 while no new calls are
any time k, each vehicle j is assigned to follow a
received. When a new service request is received, the
sequence of pickups and deliveries (control action),
dispatcher calculates the control sequence in the next
and can be represented by the function Sj 4k5, where
step S4k5 for the fleet of vehicles, adding the stops
the ith element of the sequence represents the ith stop
indicated by the new customer. Then, each sequence
of vehicle j along its route, and wj 4k5 is the total num-
Sj 4k5 remains fixed during the whole time interval
ber of stops. A stop is defined by a user who requires
4k1 k +15, unless a vehicle reaches a predefined pickup
the service (it could be its pickup or delivery). The
or delivery stop during such an interval, in which case
initial condition sj0 4k5 corresponds to the position of
its sequence will decrease in size showing that the
vehicle j at instant time k. The set of sequences
scheduled task has been accomplished. Thus, in this
S4k5 = 6S1 4k5T 1 0 0 0 1 Sj 4k5T 1 0 0 0 1 SF 4k5T 7T associated with
scheme the problem is formulated in terms of a vari-
the fleet of vehicles correspond to the control (manip-
able time step (triggered by events), which represents
ulated) variable u4k5. The sequence of stops assigned
the time interval between two consecutive requests,
to vehicle j at instant k, Sj 4k5, is given by
0
sj 4k5 Sj (k − 1) ≡ 1+ → 3+ → 1– → 2+ → 2− → 3−
s 1 4k5
j 1−
Sj 4k5 = 00
0 3+
3−
wj 4k5
sj 4k5
label1j 4k5
1
1 − rj1 4k5 âj1 4k5
rj 4k5
2+
00 00 00 00 2−
0 0 0 0
1+
r i 4k5 i i i
= j 1 − rj 4k5 âj 4k5 labelj 4k5 1 (3) vj
00 00 00 00
0 0 0 0
wj 4k5 wj 4k5 wj 4k5 wj 4k5
Figure 1 (Color online) Representation of a Sequence of Vehicle j
rj 4k5 1 − rj 4k5 âj 4k5 labelj 4k5 and Its Stops
Muñoz-Carpintero et al.: Methodology Based on Evolutionary Algorithms to Solve DPDP
242 Transportation Science 49(2), pp. 239–253, © 2015 INFORMS
that is to say, the predictive controller takes a routing i.e., Skk+N = 6S4k5T 1 0 0 0 1 S4k + N − 15T 7T , and applies the
decision when a new call enters the system. updated sequence set S4k5 based on a receding hori-
zon strategy. The optimization variables are the cur-
2.2. Predictive Dynamic Model and Objective rent sequence that incorporate the new request, and
Function
the future sequences that incorporate the prediction
In the DPDP the state space variables include the
of future requests. Thus, the objective function com-
clock time of departure Tji 4k5 and the vehicle load
prises all of the scenarios h that consists of the sequen-
Lij 4k5, after vehicle j leaves stop i, both computed at
Downloaded from informs.org by [128.122.253.212] on 23 May 2015, at 06:02 . For personal use only, all rights reserved.
ph(k + 1), H(k + 1) Demand The ad-hoc methodology for the HPC of the dial-a-
k
predictor ride system has been designed based on the particular
structure of the problem, and is presented next.
Xj (k + 1)
HPC based on Sj(k) Dial-a-ride 3.1. Methodology Based on Evolutionary
evolutionary Tj (k + 1)
system Algorithms (EA)
algorithms Lj (k + 1)
In this section, we present the developed methodol-
Downloaded from informs.org by [128.122.253.212] on 23 May 2015, at 06:02 . For personal use only, all rights reserved.
sj0 4k5
Evolutionary stage
sj1 4k5
00
0
pu
No sj 4k5
End
v
condition? sj 4k5 = 00
0
sjde 4k5
(Sub) optimal solution 00
0
wj 4k5
Figure 3 Generic Evolutionary Algorithm sj 4k5
Muñoz-Carpintero et al.: Methodology Based on Evolutionary Algorithms to Solve DPDP
244 Transportation Science 49(2), pp. 239–253, © 2015 INFORMS
00 00 00 00
Step (3). Provided that S l 4k5 is known, evaluate
0 0 0 0
the fitness function f2nd level 4k5, defined in (10), for all
wj 4k5 wj 4k5 wj 4k5 wj 4k5
rj 4k5 1−rj 4k5 âj 4k5 labelj 4k5
potential feasible sequences S m 4k + 15h . If S m 4k + 15h
The particular encodings of the insertion positions or S l 4k5 are unfeasible in terms of the capacity of
associated with each method, as well as the repair the vehicle or precedence, penalize its fitness (if a
strategies designed to handle unfeasible solutions, are repair method is considered, the unfeasible solutions
defined in §3.2. in terms of precedence constraints are repaired using
For a specific vehicle, the first level considers the ObtainSequence; thus, at this point the sequence is fea-
insertion positions of the incoming call; the fitness sible, at least in precedence).
function for the first level f1st level 4k5 is given by the Step (4). If an ending criterion (maximum num-
objective function associated with the whole opti- ber of iterations) is satisfied, then proceed to Step 5,
mization problem, as defined in (7). As this objective and return best solutions for every probable request.
function considers the insertion costs of the expected Otherwise, use the function EvolutionaryStage to
future requests, the second level of the algorithm update the particles or individuals. By using the
must be run inside the objective function compu- function ObtainSequence, the sequences S m 4k + 15h ,
tation step of the first level in order to find near- m2 11 21 0 0 0 1 n for h2 11 21 0 0 0 1 H 4k + 15 are generated
g2 = g2 + 1. Go back to Step 3.
optimal insertion positions for every expected future
Step (5). At this step, the second level ends and we
request, assuming that the candidate solution for the
go back to the first level. Given that S l 4k5 is known
incoming request being evaluated at the first level is
and the solutions for S m 4k +15h , h2 11 0 0 0 1 H 4k +15 are
assigned to such a vehicle. In addition, as the sec-
obtained in Step 3, the objective function f1st level 4k5,
ond level considers only the insertion positions of
in this case considering two steps ahead, is evaluated
one of the expected requests (h) provided a candidate
and used as the fitness function. If S l 4k5 is unfeasible,
solution for the incoming request S l 4k5, the second
penalize it.
level fitness, namely, f2nd level 4k5, considers the follow-
Step (6). If a termination criterion (maximum num-
ing objective function for every candidate solution:
ber of iterations of Steps 2–6) is satisfied, then STOP,
F
X and return the best solution found. Otherwise, by
f2nd level 4k5 = 4Cj 4k + 25 − Cj 4k + 155S l 4k51 h 0 (10) using the function EvolutionaryStage, the positions and
j=1 the velocities of all particles for l2 11 21 0 0 0 1 n are
updated. Using ObtainSequence, S l 4k5, l2 11 21 0 0 0 1 n,
Next, we describe the two-level HPC-EA-DRS algo- are generated g1 = g1 + 1. Go back to Step 2.
rithm that uses modules (see Figure 3) that in general Next, we present the different evolutionary algo-
will be different depending on the chosen evolution- rithm strategies, which are used as a part of this ad-
ary algorithm (these modules are described in Online hoc algorithm.
Appendix B (available as supplemental material at
http://dx.doi.org/10.1287/trsc.2014.0569). 3.2. Evolutionary Algorithms for the
The HPC-EA-DRS is as follows: HPC-EA-DRS
Step (0). Call function InitializeParameters of the The evolutionary algorithms used in the HPC-EA-
selected evolutionary algorithm. DRS are described in this section, including its basics
Step (1). Suppose that the predefined sequence set and the actual operators according to the HPC formu-
S4k − 15 is known. A new service request (call) enters lation for the DPDP.
the system. The first level of the algorithm then starts First, we highlight the PSO algorithm, which is
and the first level counter is set to g1 = 0. The modules based on a particle swarm that represents a popu-
InitPopulation and ObtainSequence are used to generate lation of candidate solutions (Kennedy and Eberhart
a set of n potential sequences S l 4k5, with l2 11 21 0 0 0 1 n. 2001). The particles are initialized randomly, and then
Note that 6n/F 7 candidate solutions are associated they move iteratively within the search space in order
with each vehicle, which means that the insertion of to find new solutions. The particles have a fitness asso-
the new call falls in the specific vehicle sequence (F is ciated with the solution quality, usually given by the
the fleet size). objective function to be optimized. Each particle i is
Muñoz-Carpintero et al.: Methodology Based on Evolutionary Algorithms to Solve DPDP
Transportation Science 49(2), pp. 239–253, © 2015 INFORMS 245
characterized by a position, a velocity, its best previ- For instance, if we use a real-valued particle rep-
ous position, and the best position among all of the resentation for vehicle j, a particle is encoded as x =
particles belonging to the swarm. The particles are 4x1 1 x2 5 = 40071 3085. To decode such a particle, we sim-
updated (they move) according to their cognitive and ply approximate each coordinate to the upper integer
social behavior. The above description of PSO was to obtain the insertion positions = 4pu1 de5 = 411 45.
originally conceived to solve continuous problems. In In this paper, three strategies are used for dealing
this work, we adapted the standard PSO configura- with solutions that do not satisfy the precedence con-
Downloaded from informs.org by [128.122.253.212] on 23 May 2015, at 06:02 . For personal use only, all rights reserved.
tion in order to add integer variables in the solution. straint (pickup before delivery) that appear by the
The second solver we explore is GA (Man, Tang, generation of individuals/particles of GA and PSO:
and Kwong 1998). It is based on biological evolu- —Penalty approach (called P ). In this case, a
tion, and uses inherited operators such as mutation, penalization term is added to the objective function
selection, and crossover. In particular, the optimiza- when some solutions (individuals/particles) generate
tion variables in the HPC formulation of DPDP are unfeasible sequences. Specifically, a hard penalization
discrete, and therefore the binary encoding is not nec- is used by including a very high weighing factor
essary. In other words, the genes of the individuals in computing the fitness of the unfeasible solutions,
(feasible solutions) are given directly by the integer regardless of the distance to the feasible region.
optimization variables. In addition, gradient compu- —Repair strategy (R1). The Baldwinian evolution
tations are not necessary as in conventional nonlinear (Hinton and Nowlan 1987) is considered to repair
optimization solvers, which saves significant compu- unfeasible individuals just for evaluation purposes,
tation time. Pseudocodes of both GA and PSO algo- although they are not modified in the population.
rithms are presented in Online Appendix C. Then, the population is a mix of feasible and unfeasi-
Based on the standard PSO and GA methods, nine ble individuals.
different cases are considered in this work, depend- —Repair strategy (R2). Lamarckian evolution
ing on the encoding and constraint handling methods. (Ackley and Littman 1994) is used to repair an unfea-
Two of these cases were previously proposed in Sáez, sible solution into a feasible one, which replaces the
Cortés, and Núñez (2008) and Cortés et al. (2009), and original unfeasible option in the population. Then, all
the other seven are new ad-hoc implementations. members of the population are feasible.
In these algorithms, each member of the popula- The details of the entire repair procedure are
tion represents a candidate solution for the insertion presented in Online Appendix B, function Obtain-
positions of a request for a given vehicle. Each parti- Sequence. Candidate solutions that do not satisfy the
cle or individual (particle and individual are the typ- capacity constraint are always penalized.
ical nomenclature for candidate solutions in the PSO According to the encoding and handling options of
and GA literature, respectively) has two coordinates; unfeasible solutions, six versions of PSO and three
the first represents the pickup position and the sec- versions of GA are used: P-PSO-R, R1-PSO-R, R2-
ond represents the delivery position of insertion in the PSO-R, P-PSO-N, R1-PSO-N, R2-PSO-N, P-GA, R1-
sequence of a single vehicle (no-swapping of stops is GA, and R2-GA. The prefixes P, R1, and R2 indicate
assumed as in Sáez, Cortés, and Núñez 2008; Cortés, the handling technique for unfeasible solutions based
Sáez, and Núñez 2008; Cortés et al. 2009). Particles or on penalization P, R1 for Baldwinian repair, and R2
individuals are encoded by x = 4x1 1 x2 5, which repre- Lamarckian repair, respectively. The suffixes R and N
sent a candidate solution. PSO algorithms can be used indicate that solutions are encoded in a continuous
with continuous or discrete particles whose coordi- or integer domain, respectively. R2-GA was proposed
nates are of the form x = 4x1 1 x2 5 ∈ R or x = 4x1 1 x2 5 ∈ in Sáez, Cortés, and Núñez (2008) and R1-PSO-R was
N , respectively. In the case of GA, the coordinates of proposed in Cortés et al. (2009).
each individual are directly discrete, and therefore in
such a case we considered x = 4x1 1 x2 5 ∈ N . Hereafter 3.3. Parameter Tuning for the HPC-EA-DRS
in the paper, R and N represent the sets of real and Several studies have reported the optimization of
integer numbers, respectively. parameter tuning for PSO and GA. Most of these
Recall that the insertion positions are defined as studies cover cognitive, social parameters, and iner-
= 4pu1 de5, where pu represents the position of tia weights in case of PSO; mutation, crossover rates,
the pickup, and de is the position of the delivery. and selective pressure parameters for GA. These stud-
A feasible insertion requires three conditions. First, ies offer recommended parameter sets, recognizing
the pickup and delivery positions have to be con- that these are problem dependent. Then, we perform
sistent with the length of the sequence; second, the a sensitivity analysis of these parameters; however,
pickup position in the sequence will always precede the obtained standard deviations turned out to be
the delivery position; and third, the capacity con- too large in order to draw any conclusions. Because
straint must be satisfied. of that, we decided to use a manually tuned set of
Muñoz-Carpintero et al.: Methodology Based on Evolutionary Algorithms to Solve DPDP
246 Transportation Science 49(2), pp. 239–253, © 2015 INFORMS
(0, 9) (3, 9) (6, 9) (9, 9) [Km] (0, 9) (3, 9) (6, 9) (9, 9) [Km]
3 4 3 4
Downloaded from informs.org by [128.122.253.212] on 23 May 2015, at 06:02 . For personal use only, all rights reserved.
1 2 1 2
(0, 9) (3, 9) (6, 9) (9, 9) [Km] (0, 9) (3, 9) (6, 9) (9, 9) [Km]
P\D 2 4 P\D 1 3 4
1 0, 15 0, 4 1 0, 00 0, 00 0, 23
3 0, 15 0, 3 3 0, 14 0, 12 0, 51
Although service requests are unknown, the aver- the region. The experiment is repeated 30 times for
age system pattern is assumed known from historical each case of analysis (each configuration or PSO and
data, obtained from the average demand measured GA), testing different demand requests that follow the
over the preceding week. pattern in Figure 5. The relevant indicators we mea-
In the case study, four different future requests sure are the average total cost in terms of waiting
are considered between 7:00 and 7:59 a.m., and four time, travel time, and operational cost for the entire
more between 8:00 and 8:59 a.m. Each of these two-hour period and the average computation time
future requests triggers the discrete event model at an for solving the optimization problem associated with
instant , and corresponds to the request that is opti- the insertion decision of every incoming request, as
mized during the second level optimization (two-step shown in Table 1. These tests were conducted using
ahead). Because the generated requests are just pre- 10 generations and 10 individuals in the evolution-
dictions, the calculated “near-optimal” sequences for ary algorithms used within the HPC-EA-DRS method.
such requests are not used for the dispatcher, meaning The considered configurations of evolutionary algo-
that only the sequence S4k5 for the effective current rithms are (as introduced in §3.2): P-PSO-R, R1-PSO-
call is applied to the system. R, R2-PSO-R, P-PSO-N, R1-PSO-N, R2-PSO-N, P-GA,
The demand patterns and their probabilities are R1-GA, and R2-GA.
shown in Figure 5 and were determined by the zon- Our routines for all proposed algorithms based on
ing method used in Cortés et al. (2009). PSO and GA were coded in Matlab, including the dif-
We consider an urban service area of approximately ferent configurations explained above. The equipment
81 km2 . The vehicles are assumed to travel straight utilized for the implementation of the routines is an
between stops at an average speed of 20 km/hr over iMac CPU Intel Core i3 (3.2 GHz, 4 GB).
Table 1 Simulation Statistics for the Proposed Algorithms for Control the Dial-a-Ride System
EA Mean Worst case Best case Standard Mean Worst case Best case Standard
configuration [min] [min] [min] deviation [min] [s] [s] [s] deviation [s]
Table 2 Comparison Between PSO-N and PSO-R Strategies Table 3 Comparison Between Penalty and Repair Strategies
Performance change when shifting from a PSO-N to a PSO-R strategy Performance change when shifting from a penalty to a repair configuration
7,100
Population Number of Mean computation Mean total
size generations time per request [s] cost [min]
7,000
5 5 100846 6195803
8 5 160150 6187409
Downloaded from informs.org by [128.122.253.212] on 23 May 2015, at 06:02 . For personal use only, all rights reserved.
9,800 5,350
Downloaded from informs.org by [128.122.253.212] on 23 May 2015, at 06:02 . For personal use only, all rights reserved.
9,600 5,300
9,400 5,250
9,200 5,200
0 20 40 60 80 100 0 10 20 30 40 50 60
Mean computation time tn, g [s] Mean computation time tn, g [s]
4.4. Comparison of HPC-EA-DRS with a that the distribution of the probable future requests is
Double-Horizon Heuristic unknown. The heuristic method considers the neces-
In this section, the proposed HPC-EA-DRS strategy sity of making good short-term decisions without
is compared with a state-of-the-art heuristic method experiencing adverse long-term effects. At the short-
based on a double-horizon-based strategy (Mitrovic- term stage, a static optimization is performed with
Minic, Krishnamurti, and Laporte 2004) for solving
the objective of minimizing the cost of serving the
the dynamic pickup and delivery problem. A double-
incoming request (coming up in real time). At the
horizon heuristic solves a dynamic problem assuming
long-term stage, the aim is to provide enough room
(slack) for making future insertions easier. The heuris-
Table 6 Optimal Parameters Found for R2-PSO-N tic originally is designed for solving a DPDP version
with time windows, and therefore, in the long-term
Pareto optimal set for R2-PSO-N (7 vehicles)
the goal is to maximize the slack time to make inser-
Population Number of Mean computation Mean total tion easier. In our formulation however, there are no
size generations time per request [s] cost [min] explicit time windows; in this case the most impor-
5 5 30403 10134300 tant constraint is given by the capacity in the vehicles;
5 8 50100 9189407 therefore, our aim is in providing enough room to
5 11 60546 9177506 satisfy future requests at a good level of service. The
5 14 70742 9170801
5 17 80614 9150507
8 14 130185 9150308 Table 8 Optimal Parameters Found for R2-PSO-N
8 17 140755 9147605
14 11 190352 9142803 Pareto optimal set for R2-PSO-N (11 vehicles)
14 17 240430 9138105
Population Number of Mean computation Mean total
size generations time per request [s] cost [min]
Table 7 Optimal Parameters Found for R2-GA 5 5 40648 5147309
5 8 60419 5135901
Pareto optimal set for R2-GA (7 vehicles)
5 14 90020 5134108
Population Number of Mean computation Mean total 5 17 90979 5130409
size generations time per request [s] cost [min] 8 11 110815 5130003
8 17 140610 5127006
5 5 110133 9188904 11 5 90626 5132007
5 8 200680 9158708 11 20 190112 5125101
5 11 290734 9147907 14 14 180809 5125908
8 5 180153 9160307 14 17 200083 5124209
8 14 500666 9138001 14 20 210524 5123407
11 5 250234 9148600 17 11 180791 5126908
11 14 620077 9135000 17 17 210944 5123008
14 5 300260 9145909 17 20 230492 5122908
14 8 450711 9139507 20 11 200019 5124504
14 11 570105 9137207 20 14 210923 5123403
17 5 350917 9139805 20 20 240961 5122700
Muñoz-Carpintero et al.: Methodology Based on Evolutionary Algorithms to Solve DPDP
Transportation Science 49(2), pp. 239–253, © 2015 INFORMS 251
Table 9 Optimal Parameters Found for R2-GA Table 10 Optimal Parameters Found for the Double-Horizon Heuristic
Pareto optimal set for R2-GA (11 vehicles) Parameter tuning for a double-horizon heuristic
Population Number of Mean computation Mean total Mean total cost [min]
size generations time per request [s] cost [min]
K \Lmin 3 4
5 5 130643 5138301
5 8 210363 5135906 8 7127100 7123601
Downloaded from informs.org by [128.122.253.212] on 23 May 2015, at 06:02 . For personal use only, all rights reserved.
methods).
0 0 0 ··· 0
With regard to the observed trade-off between accu-
1 1 0 0 ··· 0
racy and computation, we proposed a multiobjective −1
BL1 = 3 BL2 = 1 1 0 ··· 0
approach for tuning the parameters of any strategy 0 0 00 00
00 00
0 1×4 0 0 0 0
implementation, considering computation time and
1 1 ··· 1 0 4w 4k5+15×4w 4k5+150
performance as the conflicting objectives. The output j j
of the analysis is the set of combination of parameters Both the vehicle sequence matrix Sj 4k5 and the expected
that belong to the Pareto optimal set. This approach is load vector L̂j 4k + 15, change their dimension dynamically
applied to the R2-PSO-N and R2-GA-based solution by adding two rows when a new request occurs. There-
methods. The analysis allows comparing both strate- fore, the matrix dimensions of AL 1 BL1 1 BL2 are variable. The
gies, as well as defining the optimal parameter setting matrix BL1 is designed to remove the last two columns of the
according to a predefined criterion, such as the pres- sequence vector, which are not necessary for representing
load changes from step k to step k + 1. On the other hand,
ence of a knee point or a maximum allowed compu-
when a request is satisfied, the first row of the sequence
tation time. It is shown that PSO strategies are more is eliminated. In fact, the adaptive behavior is captured by
efficient than those based on GA, because the former these techniques of expansion and reduction of matrix size.
are faster and more accurate. We believe that PSO The vehicle departure time behavior is obtained by using
performs better than GA because the search heuris- the same methodology. Analytically,
tic of PSO is more suitable to the analyzed prob- E8Tj 4k + 15/k9 = Tˆj 4k + 15 = AT · Tj 4k5 + BT 4Sj 4k551
lem, which is clearly reflected by the fact that PSO
finds better solutions, even evaluating fewer candi- where
BT 4Sj 4k554wj 4k5+15×1 = BT2 · 4Sj 4k5 · BT1 53
dates. In addition, we tested our best implementation
against a well-known heuristic method from the lit- 1 0
··· 0
erature, obtaining much better performance from our 1 0 ··· 0
method in case of adding prediction to the optimiza- AT = 0 0
00 00
00 00 0 0
tion scheme, at the expense of a higher (although con-
trollable) computational cost. 1 0 ··· 0 4w 4k5+15×4w 4k−15+153
j j
tions are converted into an integer version are also As in the load state space model, the matrices AT 1 BT1 1 BT2
topics to be investigated in the future. change their dimensions dynamically.
Muñoz-Carpintero et al.: Methodology Based on Evolutionary Algorithms to Solve DPDP
Transportation Science 49(2), pp. 239–253, © 2015 INFORMS 253