Ant Colony System with Sparse Pheromone

Abstract

Ant colony optimization algorithm is a typical meta-heuristic algorithm, which is widely used in various combinatorial optimization problems, but its high space complexity, which has become one of the main constraints affecting the application of ACO algorithm. The pheromone matrix is one of the major storage overheads. This paper based on the analysis of typical ant colony optimization algorithms, it is observed that the pheromone matrix of ACS has very strong sparseness. Therefore, SACS is proposed, whose pheromone matrix uses triplet sparse storage. In order to solve the problem of how to deal with the number of items of the initial allocated pheromone triplet and the new non-default pheromone update, this paper proposed the method based on the small fixed storage space with different replacement policies, those are, SACS-Max, SACS-Min, SACS-Rand. Many experiments show that this method basically eliminating the storage bottleneck of the pheromone matrix.