Coordinating Cross-robot Resource for Task Allocation with Decentralized Cooperative Genetic Algorithm

This paper proposes a decentralized cooperative genetic algorithm (DCGA) to address the cross-robot resource allocation problem in leaderless multi-robot systems (MRS), where each agent is only capable of making local decisions based on local information. Inspired by the fundamental concept of divide-and-conquer in co-evolution mechanisms, the DCGA alternates between two steps: local evolution and global negotiation, within each robot. Our problem formulation enables the algorithm to effectively manage complex cross-robot constraints, which are often challenging in distributed algorithms. Unlike parallelized algorithms that necessitate agents to possess global decision-making capabilities, the DCGA facilitates the seamless integration of agents into multi-agent systems (MAS) by merely incorporating information interaction logic. Experimental results demonstrate that the DCGA is competitive in terms of optimization speed and yields superior solutions compared to centralized genetic algorithms (CGA). Furthermore, it exhibits adaptability to network latency in wireless environments.