巡回セールスマン問題に対するPSO-ACOハイブリッド手法の提案(システム情報工学科)

概要

論文の詳細を見る
This paper focuses on improving PSO to apply for TSP. First, we adopt the method of making the original PSO for TSP, using the idea of the Huang's exchange operator. Then, we confirm the optimal tuning of parameters for TSP-PSO. Although, we find out that this method is apt to get into local optimal solutions in larger scale problems, in spite of quick convergence. Next we improve the Huang's method by using GA's mutation, but the result is not satisfactory to get over this disadvantage. So, we newly propose a hybrid technique of PSO and ACO. In PSO-ACO, we consider the storage function of search information is given to a swarm besides the function that ants leave pheromone in the search place. Side-by-side tests on the problem of 150 cities between mutation type PSO, ACO and PSO* ACO hybrid technique are done. As a result, the best solution of the hybrid technique excels the other two techniques, and is a little superior to that also of other researchers' GA techniques.