An adaptive evolutionary approach for real-time vehicle routing and dispatching

The quality of the convergence process in genetic algorithms depends on the specific choice of strategies and combinations of operators. In this paper, we address this problem and introduce an adaptive evolutionary approach that uses a genetic algorithm in an adaptive process. An application of this approach to the dynamic vehicle routing problem with time windows is presented. We compare the adaptive version of a hybrid genetic algorithm with the non-adaptive one with respect to the robustness and the quality of the generated solutions. The results obtained show the ability of our operator combination adaptation approach to produce solutions that are superior to hand-tuning and other adaptive methods with respect to performance sensitivity and robustness.

► An adaptive approach for dynamic vehicle routing problem with time windows is proposed.
► Adaptation concerns the combinations of operators rather than individual operator probabilities.
► The use of the higher-level genetic algorithm instead of hand-tuned parameters and some others adaptation techniques shows an improvement over the state of the art with respect to performance sensitivity and robustness.
► Proposed approach allows different levels of parallelization to cope with real-time environments.