Rank-based ant system with originality reinforcement and pheromone smoothing

Abstract

Ant Colony Optimization (ACO) encompasses a family of metaheuristics inspired by the foraging behaviour of ants. Since the introduction of the first ACO algorithm, called Ant System (AS), several ACO variants have been proposed in the literature. Owing to their superior performance over other alternatives, the most popular ACO algorithms are Rank-based Ant System (ASRank), Max-Min Ant System (MMAS) and Ant Colony System (ACS). While ASRank shows a fast convergence to high-quality solutions, its performance is improved by other more widely used ACO variants such as MMAS and ACS, which are currently considered the state-of-the-art ACO algorithms for static combinatorial optimization problems. With the purpose of diversifying the search process and avoiding early convergence to a local optimal, the proposed approach extends ASRank with an originality reinforcement strategy of the top-ranked solutions and a pheromone smoothing mechanism that is triggered before the algorithm reaches stagnation. The approach is tested on several symmetric and asymmetric Traveling Salesman Problem and Sequential Ordering Problem instances from TSPLIB benchmark. Our experimental results show that the proposed method achieves fast convergence to high-quality solutions and outperforms the current state-of-the-art ACO algorithms ASRank, MMAS and ACS, for most instances of the benchmark.