Abstract

In recent years, there has been an increasing interest in Opposite Learning strategies. In this work, we propose COISA, a Cooperative Opposite-Inspired Strategy for Ants. Inspired on the concept of anti-pheromone, in this approach, sub-colonies of ants perform different search processes to construct an initial pheromone matrix. We aim to produce a repel effect to (temporarily) avoid components that were related to an undesirable characteristic. To assess the effectiveness of COISA, we selected Ant Knapsack, a well-known ant-based algorithm that efficiently solves the Multidimensional Knapsack Problem. Results in benchmark instances show that the performance of Ant Knapsack is improved considering the opposite information, so that it can reach better solutions than before.