A novel approach to avoiding early stagnation in Ant Colony Optimization algorithms

Abstract

This paper introduces a novel approach to avoiding early stagnation in Ant Colony Optimization algorithms. The approach involves oscillating the α and β parameters out of phase with each other according to an offline adaptation formula triggered by the online signal of stagnation in improved solutions across iterations. Further, in this paper, we present the experimental results obtained from applying this method to solving the Traveling Salesman Problem across eight fully connected, symmetric maps of sizes ranging from 51 to 1,400 cities, and show that a marginal improvement is achieved even with relatively constrained amounts of computation time and in the absence of fine-tuning of the ACO parameters towards each specific instance of the problem.