Cooperative control for swarming systems based on reinforcement learning in unknown dynamic environment

Abstract

This paper discussed the cooperative control problem for swarming systems in unknown dynamic environment. The swarm agents are required to move in a completely distributed manner with the reference trajectory determined by a virtual dynamic leader. In addition to keeping an appropriate distance from neighboring agents, each agent needs to avoid collision with dynamic threats in unknown environment. All of these complex requirements are integrated and designed as the performance index function for each agent. Then, the cooperative learning behavior of swarming system is realized by applying the reinforcement learning theory. Neural networks are used to model the control scheme and trained to minimize the performance index. The online updating rules of the neural networks are achieved based on the gradient descent algorithm. Finally, two simulation experiments are performed to verify the effectiveness of the cooperative control scheme and the environmental adaptability of the swarm agents.