Intelligent Engineering Systems through Artificial Neural Networks, Volume 16
11 Multi-Agent Evolutionary Approach to the TSP
Download citation file:
- Ris (Zotero)
- Reference Manager
A new parallel and distributed algorithm is proposed for the Traveling Salesman Problem (TSP) based upon a Multi-agent Evolutionary Algorithm (MAEA). An agent is assigned to a single city and builds locally its neighborhood - a subset of cities, which are then considered as local candidates to a global solution of TSP. The global solution of the TSP problem is based on an Ant Colonies (AC)-like paradigm. The cycles that are found by the ants are placed in data structure called Global Table, and are evaluated by genetic algorithm (GA) to modify the rank of cities in local neighborhoods. We present a description of the algorithm and show how values of some parameters of MAEA influence on the quality of solutions. We present the results of an experimental study which shows that the proposed algorithm outperforms two other currently used metaheuristics - AC and artificial immune systems.