An Evolutionary Game Coordinated Control Approach to Division of Labor in Multi-Agent Systems

In this paper, we propose an evolutionary game theoretic approach to coordinated control of multi-agent systems. In this mathematical framework, agents play games with their neighbors on the network, and update strategies through local interaction. In order to achieve a certain control objective of the system, we need to select the appropriate game type, design the calculation and evaluation methods of fitness, specify the interactive constraints and updating rules. During the evolutionary process of the system, agents have no predesigned dynamical equations. They adjust their behavior independently for the purpose of increasing their own benefits. The system achieves its final state in the process of individual interaction and autonomous decision-making. Taking division of labor problem as an example, we demonstrate the proposed control approach in detail. The performance of the theoretical method is verified by simulation on the regular graph, the general connected graphs, and heterogeneous scale-free networks, respectively.