Evolutionary Niche Artificial Fish Swarm Algorithm for Dynamic Subgroup Size Adjustment in Robot Swarms

Abstract

Adapting subgroup sizes adjustment in distributed multi-target search tasks for robot swarms presents a significant challenge. Traditional search methods struggle to dynamically adjust subgroup sizes as search conditions change under limited population information. This paper proposes a novel multi-robot cooperation approach known as the Evolutionary Niche Artificial Fish Swarm Algorithm (ENAFSA) for adapting subgroup sizes. By integrating niche AFSA with a Markov chain learning model, ENAFSA introduces an automatic learning strategy for adaptive subgroup size adjustment in multi-target search tasks within robot swarms. It leverages niche technology, combining it with a distributed version of the AFSA to simultaneously locate and search for targets. Additionally, ENAFSA incorporates a mutation mechanism that allows robots to autonomously reallocate among different subgroups, enabling decentralized changes in subgroup sizes. The mutation rate for each robot is determined by the probability transition matrix of the Markov chain model, and we employ the Markov Chain Gradient Descent (MCGD) method to optimize this transition matrix. We conduct simulation experiments to showcase the practicality of our subgroup adjustment algorithm and its effectiveness in searching for multiple targets, even when the number of robots and targets varies.