Evolutionary Games and Distributed Decision-Making in a Multi-agent System for Formation Control

In this work, the formation control based on dynamic game evolution under the mixed strategies is proposed. Different from the general formation control and collision avoidance, here agents are independent and make decisions in a distributed way. The motion states of the agent arc determined by strategy selection and evolution. We investigate how the agents accomplish the formation with less loss and fewer steps. Firstly. the fixation probabilities and proportions of different strategies under the different circumstances are analyzed. Then, we discuss the evolution processes and formation completions which vary with the collision cost according to the simulation results. Moreover, we extend our model to a larger scale, that is increasing the intensity of population in a fixed region, to study a more complex system. In our framework, we analyze how evolutionary stable strategy emerges and how formation control finishes better in the population. Our results provide some innovative ideas on achieving formation control and collision avoidance based on evolutionary game in the real world.