‘T’ for Tabu and Time Dependent Travel Time

Abstract

Even in its most basic form, the Vehicle Routing Problem and its variants are notoriously hard to solve. More often artificially intelligent algorithms are employed to provide near-optimal solutions. To be classified as “intelligent”, however, a solution strategy should be able to first analyze the environment in which the problem occurs, then solve it, and afterwards reflect on the solution process so as to improve future decision-making. Although reference is made to the entire context of the intelligence research project, this paper reports on the Tabu Search algorithm designed that catered for a problem with multiple soft time windows, a heterogeneous fleet, double scheduling, and time dependent travel time. An adaptive memory procedure was employed, initially populated with good initial feasible solutions, and the algorithm was tested on 60 problems based on established benchmark sets. The complex variant of the Vehicle Routing Problem required between 670 and 4762 seconds on a standard laptop computer, which is considered to be reasonable in the proposed application, and was consistent between different runs with an absolute mean deviation of 3.6%. The contribution is significant as it provides an algorithm that efficiently addresses a complex and practical application of the Vehicle Routing Problem. The algorithm can easily be extended to make use of multiple processors so as to reduce computational time.

Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.