GSP-ANT: An efficient ant colony optimization algorithm with multiple good solutions for pheromone update

Ant colony optimization (ACO) is a metaheuristic for various optimization problems, especially the hard combinatorial optimization problems. However, existing ACO algorithms suffer from search stagnation and exorbitantly long computation time. To alleviate these shortcomings, an improved ACO algorithm, called GSP-ANT, is presented in this paper. It maintains a good solution pool (GSP) and alternately uses the optimal solution and suboptimal solutions in the pool to update pheromone. This enables ants to transfer among different solution regions and accordingly explore larger solution space. On the other hand, once a solution in the GSP is selected, it is continuously used for pheromone update in a certain number of iterations with the aim of exploiting the neighborhood of this solution intensively. By this means, both the intensification and diversification of the search are considered. The performance of GSP-ANT is examined experimentally on typical traveling salesman problems. Computational results indicate that GSP-ANT is a promising approach.