Second order swarm intelligence

Abstract

An artificial Ant Colony System (ACS) algorithm to solve general-purpose combinatorial Optimization Problems (COP) that extends previous AC models [21] by the inclusion of a negative pheromone, is here described. Several Traveling Salesman Problem (TSP) were used as benchmark. We show that by using two different sets of pheromones, a second-order coevolved compromise between positive and negative feedbacks achieves better results than single positive feedback systems. The algorithm was tested against known NP-complete combinatorial Optimization Problems, running on symmetrical TSPs. We show that the new algorithm compares favorably against these benchmarks, accordingly to recent biological findings by Robinson [26,27], and Grüter [28] where "No entry" signals and negative feedback allows a colony to quickly reallocate the majority of its foragers to superior food patches. This is the first time an extended ACS algorithm is implemented with these successful characteristics. © 2013 Springer-Verlag.