pvh@csbrown.edu} T he vehicle routing problem with time windows is a hard combinatorial optimization problem that has received considerable attention in the last decades. This paper proposes a two-stage hybrid algorithm for this transportation problem. The algorithm first minimizes the number of vehicles, using simulated annealing. It then minimizes travel cost by using a large neighborhood search that may relocate a large number of customers. Experimental results demonstrate the effectiveness of the algorithm, which has improved 10 (17%) of the 56 best published solutions to the Solomon benchmarks, while matching or improving the best solutions in 46 problems (82%). More important perhaps, the algorithm is shown to be very robust. With a fixed configuration of its parameters, it returns either the best published solutions (or improvements thereof) or solutions very close in quality on all Solomon benchmarks. Very preliminary results on the extended Solomon benchmarks are also given.
IntroductionVehicle routing problems are important components of many distribution and transportation systems, including such examples as bank deliveries, postal deliveries, school bus routing, and security patrol services. They have received considerable attention in the past decades. This paper considers the vehicle routing problem with time windows (VRPTW). Given a number of customers with known demands and a fleet of identical vehicles with known capacities, the problem consists of finding a set of routes originating and terminating at a central depot and servicing all the customers exactly once. The routes cannot violate the capacity constraints on the vehicles and, in addition, must meet the time windows of the customers, which specify the earliest and latest times for the start of service at a customer site. A standard objective of the VRPTW problem consists of minimizing the number of routes or vehicles (primary criterion) and the total travel cost (secondary criterion). Other objective functions have been considered in various papers; for example, optimality results often focus only on the second criterion. The VRPTW problem is NP-complete (Lenstra and Rinnooy Kan 1981), and instances involving 100 customers or more are very hard to solve optimally. Indeed, very few of the Solomon benchmarks (Solomon 1987) involving 100 customers have been solved optimally (see Fisher et al. 1997 andKohl et al. 1999 for some recent results). As a consequence, local search techniques are often used to find good solutions in reasonable time.Early work in local search on the VRPTW often utilized simple heuristics or metaheuristics, and an excellent summary can be found in Gendreau et al. (1997). In recent years, the focus of local search has shifted to more complicated metaheuristics to increase the power of the earlier techniques. These include simulated annealing (Chiang and Russell 1996), tabu search (Chiang and Russell 1997, Cordeau et al. 2001, De Backer et al. 2000, Rochat and Taillard 1995, Taillard et al. 1997, genetic/evolutionary al...