Professional Documents
Culture Documents
www.palgrave-journals.com/jors
proposed. It consists in concurrently evolving two popula- and added to population Popp (p ¼ 1; 2). The process
tions of solutions subject to periodic migration in order to continues until the overlapping population outnumbers the
minimize total travelled distance using genetic operators initial population by np. Then, the np worst individuals are
combining variations of key concepts inspired from routing eliminated to maintain population size using the following
techniques and search strategies used for a time variant of individual evaluation:
the problem to further provide search guidance while
balancing intensification and diversification. Granular Evali ¼ di = maxðdm ; di Þ ð1Þ
neighbourhoods were also exploited to further reduce
computational cost. A computational experiment conducted where di is the total travelled distance related to individual i
on common benchmark problems shows the proposed and dm is the average total travelled distance over the
hybrid genetic approach to be competitive with the best- individuals forming the initial populations.
published methods. The lower the evaluation value the better the individual
The paper is outlined as follows. The main concepts of the score (minimization problem). An elitist scheme is also
proposed hybrid genetic algorithm (HGA) are first intro- assumed, in which the fittest solution individual automati-
duced. Then, the selection scheme, recombination and cally survives to the next generation.
mutation operators are presented. Concepts derived from A direct problem-specific representation is used in which
well-known heuristics such as large neighbourhood search,22 ordered lists of customers define solution routes. No explicit
route neighbourhood-based two-stage metaheuristic23 and l- solution encoding is required, making identical the pheno-
interchange mechanism24 are briefly outlined. The next type and genotype of a solution individual. Genetic
section presents an improved version of the algorithm. It operators are simply applied to a population of feasible
consists in reducing computational cost by restricting solutions rather than a population of encoded solutions.
exploration of the search space using granular neighbour- The general algorithm is specified as follows:
hoods.25 The results of a computational experiment to assess
the value of the proposed approach reporting comparative Initialization
performance to alternate methods are then presented. Repeat
Finally, some conclusions and future research directions p¼1
are briefly outlined in the last section. Repeat {evolve population Popp - new generation}
For j ¼ 1..np do
Select two parents from Popp
Hybrid genetic approach Generate a new solution Sj using
recombination and mutation operators
General description
associated with Popp
The proposed HGA-VRP algorithm aims at computing Add Sj to Popp
high-quality solutions to compete with the best-known end for
procedures under acceptable run-time conditions. It relies Remove from Popp the np worst individuals using
on the synergy between evolutionary principles and heuristic the evaluation function (1)
methods to further provide search guidance while maintain- p¼pþ1
ing a suitable balance between solution exploitation and Until (all populations Popp have been visited)
exploration. if (new best solution) then apply RC_M on best
Emphasizing genetic diversity, our approach consists in solution {cust. reordering}
concurrently evolving two populations of solutions (Pop1, Population migration {local best solutions exchange
Pop2), while exchanging a certain number of individuals across populations}
(migration) at the end of a new generation. Exclusively Until (convergence criteria or max number of
formed of feasible solution individuals, populations are generations)
evolved to minimize total travelled distance using genetic
operators based upon variations of known routing methods. The initialization phase involves the generation of initial
Whenever a new best solution emerges, a post-processing populations Pop1 and Pop2 using a random procedure to
procedure (RC_M) aimed at reordering customers is applied construct feasible solution individuals. Solutions are gener-
to further improve its solution quality. The RC_M mutation ated using a sequential insertion heuristic in which customers
operator is introduced in a subsequent section. The are inserted in random order at randomly chosen insertion
evolutionary process is repeated until a predefined stopping positions within routes. This strategy is fast and simple while
condition is met. ensuring unbiased solution generation. The migration
The proposed steady-state genetic algorithm resorts to consists in exchanging local best individuals from one
overlapping populations to ensure population replacement population to another. Convergence is assumed to occur
for Pop1 and Pop2. At first, new individuals are generated either when solution quality fails to significantly improve
1256 Journal of the Operational Research Society Vol. 54, No. 12
over a consecutive number of generations or, after a improved child solution using a variant of a well-known
maximum number of generations. customer insertion heuristic. The idea is to improve as much
The proposed technique is significantly different from the as possible some routes from one parent using ‘‘good genes’’
algorithm presented by Berger et al12 in many respects. The (good customer sequences) from the other. The insertion-
new procedure focuses on a single objective, deals with a based IB_X(k) crossover operator creates an offspring by
different constraint set, handles feasible solutions only, and combining, one at a time, k routes (R1) of parent solution P1
presents a simpler problem-solving method involving new with a subset of customers, formed by nearest-neighbour
and more efficient operators. routes (R2) in parent solution P2. Routes defining R1 are
determined using a randomly selected criterion out of three.
Accordingly, routes are chosen randomly following a
Selection
uniform probability distribution, or selected with a prob-
The selection process consists in choosing two individuals ability proportional either to the number of customers
(parent solutions) within the population for mating pur- characterizing a tour, or to the average distance separating
poses. The selection procedure is stochastic and biased consecutive route members. Child route generation follows
toward the best solutions using a simple roulette-wheel an iterative process. The neighbourhood R2 associated to a
scheme.8 In this scheme, the probability to select an given route r1AR1 is first created. R2 includes the routes of
individual is proportional to its fitness value. Individual P2 whose centroid is located within a certain range of r1
fitness for both populations Pop1 and Pop2 is computed as (centroid). A route centroid corresponds to a virtual site
follows: whose coordinates refer to the average position of its specific
routed customers. The range characterizing R2 is given by
fitnessi ¼ di ð2Þ average distance separating r1 from the routes defining P2. A
stochastic removal procedure is then carried out to remove
The notation is the same as in Equation (1). Better from r1, customers likely to be migrated to alternate routes.
individuals show a shorter total travelled distance (mini- Targeted customers are either selected according to waiting
mization problem). times, distance separating them from their immediate
neighbours, or randomly. Then, using a modified insertion
heuristic inspired from Solomon26 a feasible child tour is
Genetic operators
constructed, expanding the altered route r1 by inserting
Genetic operators incorporate and combine key feature customer visit candidates derived from the nearest-neigh-
variations of efficient routing techniques successfully applied bour routes R2 defined earlier. The proposed insertion
for the VRP with time windows,1 providing search guidance technique consists in adding a stochastic feature to the
through promising regions of the solution space while standard customer insertion heuristic I1,26 by randomly
reducing computation time. Even at the cost of making using cost function parameters and selecting the next
these specialized operators hardly reusable for other customer visit over the three best candidates with a bias
problems, the potential gain in performance may be quite toward the best. Once the construction of the child route is
noticeable. Procedure variations include Solomon’s inser- completed, and reinsertion is no longer possible, a new route
tions heuristic I1,26 in which non-deterministic contributions construction cycle is initiated. The overall process is repeated
have been added to both the insertion cost function and the for the k routes of R1. Finally, the child inherits the
customer selection process to further intensify exploration; remaining ‘‘diminished’’ routes (if any) of P1. If unvisited
large neighbourhood search,22 adding non-determinism to customers still remain, additional routes are built using a
the customer reinsertion process and, making the customer nearest-neighbour procedure.
visit ordering scheme less myopic by using additional The whole process is then reiterated to generate a second
information to better handle multiple insertion opportu- child interchanging P1 and P2. IB_X(k) can be outlined as
nities; and the l-interchange mechanism24 in which customer follows:
sets are exchanged over alternate neighbouring routes
exploiting a new route neighbourhood structure inspired Step 1 Select a set R1 of k routes from P1 to be visited.
from the RNETS23 metaheuristic to further reduce search. Step 2
Details on recombination and mutation operators are given Repeat
in the next sections. 2.1 Select randomly an unvisited route r1 A R1
(chromosome segment) from P1 (chromosome).
2.2 Select a subset of routes R2 from P2 located in
Recombination
the neighbourhood of r1. Customers forming the
A single recombination operator is considered, namely routes of R2 define a new set NC2.
IB_X(k). Its contribution is to build upon promising 2.3 Remove from r1 customers likely to be migrated
segments of two parent solutions to generate a new to alternate routes (large waiting times or
J Berger and M Barkaoui—Genetic alogorithm for capacitated VRP 1257
distance separating neighbour route members) or For diversification purposes, two customer reinsertion methods
randomly. are proposed, one of them being randomly selected (50%
2.4 Build a child route solution: apply a stochastic probability) on mutator invocation.
version of the insertion-based routing technique The first reinsertion method relies on the insertion cost
I1 in which the modified route r1 is the initial function prescribed by Solomon’s procedure I126 for the
child route solution and customers to be VRP with time windows and, a rank-based customer visit
S
reinserted run over NC2 U. U refers to the set ordering scheme. Customer insertion cost is defined by the
of unrouted but already visited customers. sum of key contributions referring, respectively, to increased
2.5 Inserted customers are then removed from the travelled distance and delayed service time. As for customer
remaining unvisited routes of R1. Update U. ordering, customers ({c}) are sorted (CustOrd) according to
Until (all routes r1AR1 have been visited). a composite ranking, departing from the myopic scheme
Step 3 Inherit the remaining unvisited routes of P1 (if originally proposed by Shaw. The ranking is defined as an
any) while eliminating customers already additive combination of two separate rankings, previously
routed. achieved over best insertion costs (RankCost(c)) on the
Step 4 For the remaining unrouted customers U (if one hand, and number of feasible insertion positions
any), build new routes using the nearest- (Rank|Pos|(c)) on the other hand:
neighbour procedure.
CustOrd SortðRankCost ðcÞ þ RankjPosj ðcÞÞ ð3Þ
Mutation The smaller the insertion cost (short total distance, travelled
time) and the number of positions (opportunities), the better
A suite of four mutation operators is proposed, namely
(smaller) the ranking. The next customer to be visited within
LNSB_M(d), EE_M, IEE_M and RC_M(I). Each mutator
the search process is selected according to the following
is briefly described next.
expression:
The LNSB_M(d) (large neighbourhood search based)
mutation operator relies on the concept of the large customer CustOrd½INTEGERðLrand D Þ ð4Þ
neighbourhood search (LNS) method proposed by Shaw.22
The contribution of this operator consists in mutating a where L is the current number of customers to be inserted,
solution individual by conducting a more extensive search rand is a real number over the interval [0,1] (uniform random
(larger scale) of the solution space, while directing and number generator), and D is the parameter controlling
confining exploration to most promising regions, and keeping determinism. If D ¼ 1 then selection is purely random
computation to an acceptable cost. LNS consists in exploring (default: D ¼ 15).
the search space by repeatedly removing related customers Customer position selection (value ordering) is then based
and reinserting them using constraint-based tree search on cheapest insertion cost.
(constraint programming). Customer relatedness defines a The second reinsertion method involves features of the
relationship linking two customers based upon specific successful insertion heuristic proposed by Liu and Shen23 for
properties (eg proximity and/or identical route membership), the VRP with time windows, relying on the maximization of
such that when both customers are considered simultaneously a regret insertion cost function, which concurrently takes
for a visit, they can compete with each other for reinsertion into account multiple insertion opportunities (regret cost), to
creating new opportunities for solution improvement. There- determine customer visit ordering. The regret cost based
fore, customers close to one another naturally offer inter- customer visit ordering scheme is specified as follows. In the
change opportunities to improve solution quality. Similarly, insertion procedure proposed by Liu and Shen,23 route
solution number of tours is more likely to decrease when neighbourhoods associated to unvisited customers are
customers sharing route membership are removed all repeatedly examined for customer insertion. This new route
together. As stated in Shaw,22 a set of related customers is neighbourhood structure relates one or multiple routes to
first removed. The reinsertion phase is then initiated. individual customers. In our approach, customer route
The proposed customer reinsertion technique differs from neighbourhood is strictly bounded to two tours, comprising
the procedure introduced by Shaw22 resorting to alternate routes whose distance separating their centroid from the
insertion cost functions and, customer visit ordering schemes customer location is minimal. Each feasible customer
(variable ordering scheme), using new information to carry out insertion opportunity is explored over its entire route
large neighbourhood search. Customer visit ordering deter- neighbourhood. The next customer visit is selected by
mines the effective sequence of customers to be consecutively maximizing the so-called regret cost function that accounts
visited while exploring the solution space (search tree expan- for multiple route insertion opportunities:
sion). It is believed that exploiting additional information X
already available from computation is likely to guide search Regret Cost ¼ fCc ðrÞ Cc ðr
Þg ð5Þ
toward promising regions of the solution space more efficiently. r2RNðcÞ
1258 Journal of the Operational Research Society Vol. 54, No. 12
values, returning the best solution generated shall an operators, customers subject to permutation can be swapped
improved one emerge. RC_M(I) can be outlined as follows: only if the distance separating each other is less than d*.
and BB for HGA-VRP. The results are expressed in terms of Graphics 100 MHz for RT, Silicon Graphics 4D/35 for TA
total travelled distance. Published competing methods with an and Pentium 400 MHz for BB, respectively. Explicit results
average performance gap exceeding about 1% (over all for RT and TA have been omitted because no run-time was
instances) of the best-known result, and/or failing to specify provided. It is worth noticing that reported results for WH
run-time and computational resource characteristics, or includes the best computed solution over five execution runs
reporting prohibitive run-time have been deliberately omitted as well as related cumulative run-time.
for comparison purposes. Additional results involving other The results shown in Table 1 do not show any conclusive
techniques including classical heuristics may nonetheless be evidence to support a dominating heuristic over the others.
found in Cordeau et al.9 However, solution quality and run-time reported for BB
Computational results for all problem data sets are proves the HGA-VRP method to be competitive in
summarized in Table 1, similarly to the study of Cordeau comparison to alternative techniques as it mostly matches
et al.9 The first column describes the various instances and the performance of best-known heuristic routing procedures.
their related size, whereas the second specifies total travelled Accordingly, the average solution quality deviation (0.48%)
distance and run-time (in minutes). The following columns and reasonable run-time obtained certainly shows that
refer to particular problem-solving methods. Best-known HGAs can be comparable to tabu search techniques.
results (bold values) are depicted in the last column Evaluating HGA-VRP over a larger statistical sample
(Taillard35 and, Rochat and Taillard34 for instances 5 and involving 10 simulation runs for each data set provided
10). The last row refers to average run-time and performance further evidence of that claim. Results reported in Table 2
deviation from the best-known solutions over all problem indicate average performance, standard deviation and,
instances. Related computer platforms include VAX 8600 relative performance gaps (maximum, average, minimum)
for OS, Silicon Graphics 36 MHz for GHL, Sun Ultrasparc from best-known solutions for each data set, respectively, as
10 (440 MHz) for CGL, Pentium PC 200 MHz for TV, Sun well as overall average deviation. Accordingly, the results
4/630 MP for WH, Sun Sparc4 IPC for RR, Silicon show fairly small overall deviations from best-known
1 (50) Dist 524.61 524.61 524.61 524.61 524.61 524.61 524.61 524.61
(min) 1.90 6.0 4.57 0.81 20.0 1.05 2.00
2 (75) Dist 844 835.77 835.45 838.60 835.8 835.32 835.26 835.26
(min) 0.84 53.8 7.27 2.21 50.0 43.38 14.33
3 (100) Dist 838 829.45 829.44 828.56 830.7 827.53 827.39 826.14
(min) 25.72 18.4 11.23 2.39 145.0 36.72 27.90
4 (150) Dist 1044.35 1036.16 1038.44 1033.21 1038.5 1044.35 1036.16 1028.42
(min) 59.33 58.8 18.72 4.51 285.0 48.47 48.98
5 (199) Dist 1334.55 1322.65 1305.87 1318.25 1321.3 1334.55 1324.06 1291.45
(min) 54.10 90.9 28.10 7.50 480.0 77.07 55.41
6 (50) Dist 555.43 555.43 555.43 555.43 555.4 555.43 555.43 555.43
(min) 2.88 13.5 4.61 0.86 30.0 2.38 2.33
7 (75) Dist 911.00 913.23 909.68 920.72 911.8 909.68 909.68 909.68
(min) 17.61 54.6 7.55 2.75 45.0 82.95 10.5
8 (100) Dist 878.00 865.94 866.38 869.48 878.0 866.75 868.32 865.94
(min) 49.99 25.6 11.17 2.90 165.0 18.93 5.05
9 (150) Dist 1184.00 1177.76 1171.81 1173.12 1176.5 1164.12 1169.15 1162.55
(min) 76.26 71.0 19.17 5.67 345.0 29.85 17.88
10 (199) Dist 1441.00 1418.51 1415.40 1435.74 1418.3 1420.84 1418.79 1395.85
(min) 76.02 99.8 29.74 9.11 535.0 42.72 43.86
11 (120) Dist 1043.00 1073.47 1074.13 1042.87 1043.4 1042.11 1043.11 1042.11
(min) 24.07 22.2 14.15 3.18 275.0 11.23 22.43
12 (100) Dist 819.59 819.56 819.56 819.56 819.6 819.56 819.56 819.56
(min) 14.87 16.0 10.99 1.10 95.0 1.57 7.21
13 (120) Dist 1547.00 1573.81 1568.91 1545.51 1548.3 1550.17 1553.12 1541.14
(min) 47.23 59.2 14.53 9.34 510.0 1.95 34.91
14 (100) Dist 866.37 866.37 866.53 866.37 866.4 866.37 866.37 866.37
(min) 19.60 65.7 10.65 1.41 140.0 24.65 4.73
Average deviation from best 1.03% 0.86% 0.69% 0.64% 0.63% 0.55% 0.48%
Average time (min) 33.60 46.8 13.75 3.84 222.85 24.65 21.25
J Berger and M Barkaoui—Genetic alogorithm for capacitated VRP 1261
1 (50) 524.61 0 0% 0% 0%
2 (75) 840.07 3.63 1.38% 0.57% 0%
3 (100) 830.027 1.97 0.91% 0.47% 0.15%
4 (150) 1045.21 4.09 2.21% 1.63% 0.91%
5 (199) 1327.16 7.39 3.78% 2.76% 1.87%
6 (50) 555.43 0 0% 0% 0%
7 (75) 909.68 0 0% 0% 0%
8 (100) 868.49 3.10 1.03% 0.29% 0%
9 (150) 1176.44 6.56 1.97% 1.19% 0.25%
10 (199) 1422.71 3.93 2.29% 1.92% 1.45%
11 (120) 1047.54 2.02 0.75% 0.52% 0.09%
12 (100) 819.56 0 0% 0% 0%
13 (120) 1553.95 3.87 1.32% 0.83% 0.38%
14 (100) 866.37 0 0% 0% 0%
Table 3 Comparison of basic and improved HGA-VRP Impact on computational cost was further revealed when
Problem Performance BB BB* Best similar simulations were carried out to assess HGA-VRP*
instance (n) time performance. In that setting, b ran into the interval [1.25, 2]
with Db ¼ 0.25. Results comparing the basic algorithm (BB)
1 (50) Dist 524.61 524.61 524.61 to the improved version (BB*) are reported in Table 3 for the
(min) 2.00 0.20.
2 (75) Dist 835.26 835.83 835.26 same VRP benchmark28 for a single simulation run.
(min) 14.33 13.01 Generally, the use of granular neighbourhoods proves the
3 (100) Dist 827.39 829.44 826.14 algorithm to converge more quickly toward good solutions.
(min) 27.90 15.50 For a near similar quality solution, computational results
4 (150) Dist 1036.16 1037.79 1028.42 show an average run-time reduction on the order of 33%,
(min) 48.98 42.20
5 (199) Dist 1324.06 1321.38 1291.45 representing a significant computational gain. For some
(min) 55.41 35.42 instances, run-time savings can reach up to 50%.
6 (50) Dist 555.43 555.43 555.43
(min) 2.33 1.00
7 (75) Dist 909.68 909.68 909.68
(min) 10.5 7.20 Conclusion
8 (100) Dist 868.32 868.97 865.94
A hybrid genetic algorithm (HGA-VRP) to address the
(min) 5.05 4.35
9 (150) Dist 1169.15 1171.66 1162.55 classical capacitated VRP was presented. Focusing on total
(min) 17.88 7.40 travelled distance minimization, HGA-VRP concurrently
10(199) Dist 1418.79 1420.43 1395.85 evolves two populations of solutions in which respective best
(min) 43.86 44.95 individuals are mutually exchanged through migration over
11(120) Dist 1043.11 1043.18 1042.11
(min) 22.43 10.62 each generation. Genetic operators were designed to
12(100) Dist 819.56 819.56 819.56 incorporate and combine variations of key concepts emer-
(min) 7.21 3.98 ging from recent promising techniques for a time variant of
13(120) Dist 1553.12 1550.56 1541.14 the problem, to further provide search guidance while
(min) 34.91 9.35 maintaining a suitable balance between intensification and
14(100) Dist 866.37 866.37 866.37
(min) 4.73 3.20 diversification. An improved version of the algorithm has
shown to reduce computational cost significantly by further
Average deviation from best 0.48% 0.52% confining exploration of the solution space. Results from a
Average time (min) 21.25 14.17 limited computational experiment showed that HGA-VRP is
cost-effective and very competitive in comparison to the
solutions, 0.73% on average, and around 0.36% (1%) at best-known VRP metaheuristics.
best (at worst). The algorithm also shows some relative Future work will be conducted to improve the proposed
stability when considering the standard deviation average algorithm. Alternative metaheuristic features will be exam-
ratio distribution over data sets (0.7% at worst). Statistical ined to enhance genetic operators while reducing computa-
results from Table 2 provide a new basis for comparative tional cost. Additional improvements might lie on the
and statistical analysis to study future methods. introduction and combination of various selection and
1262 Journal of the Operational Research Society Vol. 54, No. 12
population replacement schemes and, new fitness models. windows. In: Eshelman LJ. (ed). Proceedings of the 6th
Application of the approach to related problems will be International Conference on Genetic Algorithms. Morgan Kauf-
mann, San Francisco, pp 536–543.
explored as well.
19 Blanton JL and Wainwright RL (1993). Multiple vehicle routing
with time and capacity constraints using genetic algorithms. In:
Forrest S. (ed). Proceedings of the 5th International Conference
References on Genetic Algorithms. Morgan Kaufmann, San Francisco, pp
452–459.
1 Toth P and Vigo D (2002) (eds). The Vehicle Routing Problem. 20 Sangheon Han and Yoshio Tabata (2002). A hybrid genetic
SIAM Monographs Discrete Mathematics and Applications. algorithm for the vehicle routing problem with controlling lethal
Society for Industrial and Applied Mathematics, Philadelphia, gene. Int J Asian Pacific Mngt Rev 7: 287–298.
USA. 21 Peiris P and Zak SH (2000). Solving vehicle routing problem
2 Gendreau M, Laporte G and Potvin J-Y (1998). Metaheuristics using genetic algorithms. Annual Research Summary—Part I—
for the Vehicle Routing Problem, Les Cahiers du GERAD, G-98– Research, Section 1—Automatic Control, School of Electrical
52. Montreal, Canada. and Computer Engineering, Purdue University, http://www.
3 Gendreau M, Laporte G and Potvin J-Y (1997). Vehicle ece.purdue.edu/ECE/Research/AR S/ARS2000/PART_I/Sec-
routing: Modern heuristics. In: Aarts E and Lenstra JK (eds). tion1/1_19.whtml.
Local Search in Combinatorial Optimization. Wiley, Chichester, 22 Shaw P (1998). Using constraint programming and local search
pp 311–336. methods to solve vehicle routing problems. In: Maher M and
4 Glover F (1986). Future paths for integer programming and Puget J-F (eds.). Principles and Practice of Constraint Program-
links to artificial intelligence. Compu Opns Res 13: 533–549. ming, Lecture Notes in Computer Science. Springer-Verlag, New
5 Glover F and Laguna M (1997). Tabu Search. Kluwer York, pp 417–431.
Academic Publishers: Boston. 23 Liu F-H. and Shen S-Y (1999). A route-neighbourhood-based
6 Holland JH (1975). Adaptation in Natural and Artificial Systems. metaheuristic for vehicle routing problem with time windows.
University of Michigan Press: Ann Arbor. Eur J Opl Res 118: 485–504.
7 Jong De KA (1975). An analysis of the behavior of a class of 24 Osman IH (1993). Metastrategy simulated annealing and tabu
genetic adaptive systems. PhD dissertation, University of search algorithms for the vehicle routing problem. Ann Ops Res
Michigan, USA. 41: 421–451.
8 Goldberg DE (1989). Genetic Algorithms in Search, Optimiza- 25 Toth P and Vigo D (1998). The granular tabu search and its
tion, and Machine Learning. Addison-Wesley: New York. application to the vehicle routing problem. Technical Report OR/
9 Cordeau J-F, Gendreau M, Laporte G, Potvin J-Y and Semet F 98/9 DEIS, University of Bologna, Bologna, Italy.
(2002). A guide to vehicle routing heuristics. J Opl Res Soc 53: 26 Solomon MM (1987). Algorithms for the vehicle routing and
512–522. scheduling problems with time window constraints. Ops Res 35:
10 Bräysy O and Gendreau M (2004). Vehicle routing problem 254–265.
with time windows, Part II: metaheuristics to appear in 27 Harvey WD and Ginsberg ML (1995). Limited discrepancy
Transportation Sc. http://www.sintef.no/static/am/opti/pro- search. In: Mellish CS (ed). Proceedings of the 14th IJCAI
jects/top/. Montreal, Canada. Morgan Kaufmann, San Fransisco, pp 607–
11 Dalessandro SV, Ochi LS, de A and Drummond LM (1999). A 615.
parallel hybrid evolutionary metaheuristic for the period vehicle 28 Christofides N, Mingozzi A and Toth P (1979). The vehicle
routing problem. In: Rolim JDP et al (eds). Proceedings of the routing problem. In: Christofides N, Mingozzi A, Toth P and
IPPS/SPDP 1999, 2nd Workshop on Biologically Inspired Sandi C (eds). Combinatorial Optimization. Wiley: Chichester,
Solutions to Parallel Processing Problems, San Juan, Puerto pp 315–338.
Rico, Springer, pp 183–191. 29 Wall M (1995). GAlib—A C þ þ Genetic Algorithms Library,
12 Berger J, Barkaoui M and Bräysy O (2003). A route-directed version 24. http://lancet.mit.edu/galib-2.4/. MIT: Boston.
hybrid genetic approach for the vehicle routing problem with 30 Gendreau M, Hertz A and Laporte G (1994). A tabu search
time windows. Inform Systems Opl Res 41: 179–194. heuristic for the vehicle routing: problem. Mngt Sci 40: 1276–
13 Machado P, Tavares J, Pereira F and Costa E (2002). Vehicle 1290.
routing problem: doing it the evolutionary way. In: Langdon 31 Cordeau J-F, Gendreau M and Laporte G (1997). A tabu search
WB et al (eds). Proceedings of the Genetic and Evolutionary heuristic for the periodic and multi-depot vehicle routing
Computation Conference New York, USA. Morgan Kaufmann, problems. Networks 30: 105–119.
San Francisco, p 690. 32 Wark P and Holt J (1994). A repeated matching heuristic for the
14 Gehring H and Homberger J (2001). Parallelization of a two- vehicle routing problem. J Opl Res Soc 45: 1156–1167.
phase metaheuristic for routing problems with time windows. 33 Rego C and Roucairol C (1996). A parallel tabu search
Asia-Pacific J Opl Res 18: 35–47. algorithm using ejection chains for the vehicle routing problem.
15 Tan KC, Lee LH and Ou K (2001). Hybrid genetic algorithms In: Osman IH and Kelly JP (eds). Meta-Heuristics: Theory and
in solving vehicle routing problems with time window con- Applications. Kluwer, Boston, pp 661–675.
straints. Asia-Pacific J Opl Res 18: 121–130. 34 Rochat Y and Taillard ED (1995). Probabilistic diversification
16 Thangiah SR, Osman IH, Vinayagamoorthy R and Sun T and intensification in local search for vehicle routing. J
(1995). Algorithms for the Vehicle Routing Problems with Time Heuristics 1: 147–167.
Deadlines. Am J Math Mngt Sci 13: 323–355. 35 Taillard ED (1993). Parallel iterative search methods for vehicle
17 Thangiah SR (1995). Vehicle routing with time windows using routing problems. Networks 23: 661–673.
genetic algorithms. In: Chambers L. (ed). Application Handbook
of Genetic Algorithms: New Frontiers. Vol. II. CRC Press, Boca
Raton, FL, pp 253–277.
18 Thangiah SR (1995). An adaptive clustering method using a Received March 2003;
geometric shape for vehicle routing problems with time accepted September 2003 after one rivision