Roach infestation optimization with friendship centers

Abstract

Roach infestation optimization (RIO) is a new adaption of particle swarm optimization (PSO) that significantly improves algorithm effectiveness in finding the global optima. This paper assesses the effectiveness of using swarm centers to further improve RIO convergence performance. Swarm centers have previously been applied in PSO as the center PSO. This paper introduces two RIO variants using one center agent and individual friendship center agents. In the first, the center agent has no explicit velocity and is positioned in each iteration at the center of the swarm. In the second, each individual friendship center adopts a position located at the center of its friends. This paper conducts experiments on 13 benchmark function optimization problems, 2 neural network learning problems, and 2 engineering design problems. Experimental results show that the RIO with a swarm center did not perform as well as the center particle in improving PSO. The behavior of Find_Friends in RIO requires each roach agent to move toward its friendship center rather than oscillate around the swarm center. The friendship centers significantly improved RIO in terms of convergence speed and stability with a minor 37.47% additional time cost.