Random Exploration and Attraction of the Best in Swarm Intelligence Algorithms

In this paper, it is revealed that random exploration and attraction of the best (REAB) are two underlying procedures in many swarm intelligence algorithms. This is particularly shown in two of the most known swarm algorithms: the particle swarm optimization (PSO) and gray wolf optimizer (GWO) algorithms. From this observation, it is here proposed that instead of building algorithms based on a narrative derived from observing some animal behavior, it is more convenient to focus on algorithms that perform REAB procedures; that is, to build algorithms to make a wide and efficient explorations of the search space and then gradually make that the best-evaluated search agent to attract the rest of the swarm. Following this general idea, two REAB-based algorithms are proposed; one derived from the PSO and one derived from the GWO, called REAB-PSO and REAB-GWO, respectively. To easily and succinctly express both algorithms, variable-sized open balls are employed. A comparison of proposed procedures in this paper and the original PSO and GWO using a controller tuning problem as a test bench show a significant improvement of the REAB-based algorithms over their original counterparts. Ideas here exposed can be used to derive new swarm intelligence algorithms.