Integrating machine learning and bee algorithms with multi-agent systems for dynamic vehicle routing problem with time windows

Abstract

This paper presents an approach to solve the Dynamic Vehicle Routing Problem with Pickup and Delivery Time Windows (DVRPPDTW) by Learning Bee Algorithm (LBA) which integrates Machine Learning (ML) with Bee Algorithm (BA) and Multi-Agent Systems (MAS). The proposed algorithm utilizes Random Forest (RF) to tune the parameters of the BA in a dynamic way enhancing its adaptability and efficiency in different real-time scenarios. MAS further improve the algorithm by enabling decentralized decision making where each vehicle act as an independent agent capable of real-time route adjustments. This hybrid approach addresses the difficulties of DVRPPDTW by optimizing routes in response to dynamic demands and conditions resulting in significant reductions in total travel distance and improvements in delivery efficiency. The proposed algorithm reduced the total travel distance by up to 5% and increased the number of deliveries by 12% in highly dynamic environments compared to existing method. The proposed method consistently outperforms existing algorithm when the performance analyzed which offer scalable and robust solution for such logistics problems. The results highlight the effectiveness of integrating ML with metaheuristics (MHs) in optimizing dynamic vehicle routing making this approach valuable contribution to the field.

Keywords

• Dynamic Vehicle Routing
• Pickup and Delivery Time Windows
• Bee Algorithm
• Machine Learning
• Multi-Agent Systems
• Random Forest