Dual Ant Colony Algorithm Based on Backtracking Migration and Matching Learning

Abstract

In order to solve the problems of slow convergence speed and easy to fall into local optimum of tradi-tional ant colony algorithm in solving traveling salesman problem (TSP), the dual ant colony algorithm based on backtracking migration and matching learning (BMACS) is proposed. Firstly, the population is dynamically graded into exploration ants and tracking ants. The ants with higher fitness are exploration ants and the ones with lower fitness are tracking ants. Secondly, a local feature transfer mechanism is proposed, in which there are two strategies: in the feature transfer strategy, the pheromone on the common path of exploration ants is rewarded, and this path is transferred into the pheromone matrix as a local feature, so as to improve the influence of exploration ants and acce-lerate the convergence speed of the algorithm; in the mutation learning strategy, the tracking ants are responsible for exploring the suboptimal path, and adaptively reconstruct the path of exploration ants to enrich the diversity of the population. Finally, when the algorithm stagnates, the matching learning mechanism is used to communicate and learn between the current optimal individual and the historical optimal individual with the highest similarity, and the pheromone is reorganized to increase the diversity of the population, thus improving the ability of the algorithm to jump out of the local optimal. Simulation experiments are carried out with MATLAB for multiple sets of cases in TSPLIB, and the results show that the improved algorithm balances diversity and convergence speed, and effecti-vely improves the quality of solution.

Keywords

• |ant colony algorithm|traveling salesman problem (tsp)|local feature migration|matching learning|dyna-mic classification