Dynamic vehicle routing by means of a genetic algorithm

Abstract

PurposeTo propose and to evaluate a new genetic algorithm (GA) for solving the dynamic pickup and delivery problem with time windows (DPDPTW).Design/methodology/approachFirst, a grouping genetic algorithm (GGA) for the (static) PDPTW is described. In order to solve the dynamic problem, the GGA then is embedded in a rolling horizon framework. Special updating mechanisms are provided which assure that reusable solution knowledge is preserved over the plan revisions. The approach is evaluated using a large number of test instances with varying degrees of dynamism.FindingsThe experimental results have demonstrated that the proposed approach is able to find high‐quality solutions when compared with two comparative heuristics.Research limitations/implicationsFuture research will be dedicated to the following issues: testing the proposed method using larger problem instances, using more sophisticated objective functions in order to further improve and evaluate the approach, integrating fast local search techniques into the genetic search, speeding up the algorithm by optimizing its implementation.Practical implicationsIn order to meet the increasing demands on the flexibility and the promptness of transportation services, algorithms are needed for dispatching transportation requests that arrive dynamically during the planning period. The findings of this contribution justify the employment of GAs in such dynamic transportation planning environments.Originality/valueAlthough the application of GAs in dynamic environments attracts growing attention, up to now no such algorithm has been published for the DPDPTW. To the best of the author's knowledge, this is the first time a GA has been applied to the DPDPTW.

Distribution of waiting time for dynamic pickup and delivery problems

“…At 50 % dod, the increase is 100.08 % for the distance and 161.89 % for the fleet size while at 100 % dod the gap to the static solution increases to 145.12 % in terms of distance and 261.56 % in terms of vehicles. The findings are consistent with the results presented by Pankratz (2005) who investigated the influence of the degree of dynamism (dod) on the solution quality and pointed out the importance of incorporating anticipatory knowledge.…”

“…However, some benchmark sets for dynamic pickup and delivery problems exist in the literature. The test setup consists of dynamic PDPTW instances that have been proposed by Pankratz (2005) which are derived from the static pickup and delivery instances provided by Li and Lim (2001) which in turn are based on the classical Solomon benchmark instances for the vehicle routing problem with time windows proposed by Solomon (1987). Each of the instances contains 50 dynamically appearing requests.…”

Optimization in dynamic environments: a survey on problems, methods and measures

“…There are also some works on dynamic vehicle routing problem made by different researchers (Hanshar and Ombuki-Berman 2002;Pankratz 2005) that could be applied to real-world problems modeled as instances of this problem. Rossi et al (2008) worked on the pose problem where there is a need to know the pose of a moving object using evolutionary algorithms and Kalman filters.…”

Visual Analytics Approach to User-Controlled Evacuation Scheduling

“…For this class of optimization problems, deterministic mathematical techniques do not work sufficiently well while heuristic methods, such as Genetic Algorithms [3], offer a powerful alternative [4,5,6]. As we have noted, the problem usually cannot be solved fully automatically and requires the involvement of a human analyst with his/her domain expertise and knowledge of the geographic area.…”