A Multi-agent Approach To Solving Dynamic Traveling Salesman Problem

Abstract

A multi‐agent approach to solving dynamic Traveling Salesman Problem (TSP) is presented. In the dynamic version of TSP cities can be dynamically added or removed. Proposed multi‐agent approach is based on the Sensitive Stigmergic Agent System model refined with new type of messages between agents. The agents send messages every time a change occurs, for instance, when an agent observes that a city has disappeared or appeared. The system is tested for various pheromone sensitivity level values and learning ability for agents. Numerical results and comparisons with the Ant Colony System and Ant System models indicate a good performance of the proposed model.