Dynamic VRP Optimization Using Discrete PSO in Edge Computing Environment

Abstract

This paper proposes using Discrete Particle Swarm Optimization (PSO) to generate an immediate solution for the dynamic vehicle routing problem (VRP). There are many applications of dynamic VRP in the real world, where not every target location to be visited is known at the very beginning, and each target location is informed at any arbitrary time. Therefore, a routing system is required to route any available vehicle to visit all target locations. Furthermore, an algorithm is needed to provide route updates when a new location is acquired. Each location must be visited within a specific time, and all vehicles must go back to the depot before the depot closing time. This variant of VRP can be categorized as dynamic capacitated VRP with time windows (DCVRPTW). An edge computing environment is designed to distribute the management of vehicles to multiple edge nodes so that the cloud server does not need to handle the vehicle, thus cutting the computation time. A discrete PSO based algorithm is designed and deployed at each edge node to solve this problem. Computer simulations show that the proposed system can route available vehicles to all target locations within their time windows with minimal routing time by edge nodes.