Optional and mandatory assignment strategies for dynamic vehicle routing with time windows

Abstract

The Dynamic Vehicle Routing Problem (DVRP) extends the classical Vehicle Routing Problem (VRP) by incorporating real-time updates, requiring continuous adjustments to routing plans. This study introduces two novel solution strategies for the DVRP with Hard Time Windows: Multi-Planning with Acception/Rejection Policy (MPA) and Multi-Planning with Mandatory Assignment Policy (MPM). Both approaches consider previously assigned, new, and simulated future customer requests, enabling more informed decision-making under dynamic conditions. In MPA, customer requests can be rejected if future requests are expected to be more advantageous, whereas MPM mandates accepting requests provided vehicle capacity is sufficient. After optimizing the methods through sensitivity analyses, we evaluated their performance against Clarke and Wright's Savings Algorithm (CW) across various real-world-inspired scenarios, including different time window lengths, interarrival times, and region types. Results show that MPA outperforms under tighter time windows and high-demand settings, while MPM demonstrates robust performance across broader conditions. The proposed methods increase the average acceptance rate by up to 81% and reduce average daily travel times by up to 48% compared to CW. Even when all methods achieve a 100% acceptance rate, MPA and MPM reduce travel times by 13% to 43%. Additionally, execution time analysis shows that shorter interarrival times and higher available vehicles significantly increase computational effort, with execution times rising up to ninefold. In contrast, variations in time window constraints have a more moderate impact, typically increasing execution times by 10% to 60%.