Modifying Ant Colony Optimization

Ant colony optimization (ACO) allows fast near optimal solutions to be found. It is useful in industrial environments where computational resources and time are limited. Like several search methods, being trapped in local optima within a search space currently is still nontrivial difficulty for ACO. In order to relieve such a difficulty, this paper focuses on modifying ant colony optimization by performing a couple of operations: i) during runtime, adjusting the parameter Q₀, which relates to exploitation and exploration in the local search of ACO. ii) immediately getting away from the trap by reinitializing Q₀ in a way of exploration. Modifying ACO by using genetic algorithm (GA) is one of several ways to accomplish i) and ii). Such a modification is called here ACO-GA. As well as GA, particle swarm optimization (PSO) could be applied in the modification of ACO in a similar manner. The latter modification is referred as ACO-PSO. The comparison is made using the NP-hard Traveling Salesperson Problem (TSP) which represents the fundamental principle of such optimizations as cost and time minimization in several industrial applications. The results show that the modifications of ACO surpass over the traditional ACO in terms of accuracy rate and iteration utilization. In the average of the TSP problems applied here, ACO-PSO generates 39.163% better than ACO-GA regarding the iteration consumption. One of the experiments also indicates that reinitializing Q₀ may not help improving the solution. The evolution of Q₀ is well enough to find the near optimum. In addition, it is also found that for the large number of nodes, the high exploration is required for achieving the minimum fitness or the least total distances.