A quickly convergent continuous ant colony optimization algorithm with Scout Ants

Many studies on ants behavior have demonstrated that their food searching process starts with Scout Ants’ scouting all around for food. In this paper, we propose a novel Scout Ant Continuous Optimization (SACO) algorithm which can simulate the food searching process of the Scout Ants. In this algorithm, the solution space of an optimization problem is divided into m subspaces. One Scout Ant is assigned to each subspace, and a total number of m Scout Ants in m subspaces will cooperate in the whole food searching process. The location of a Scout Ant in a subspace corresponds to a solution in that subspace. When an ant moves, the change of its position is driven by two factors. The first factor is the independent, random ergodic search with a small moving step in the ant’s assigned subspace, and the second is the collaborative, global search inspired by the global heuristic information accumulated among m ants. Each of these two factors is weighed by an appropriate weight to balance its contribution to the moving step size. This balanced computation helps adapt optimization problems with different features. Our numerical experiments have demonstrated that, in addition to the high accuracy and success rate in seeking the optimized solutions, our algorithm also has very fast convergence speed and impressive performance in optimization applications in high-dimensional spaces.