Prescribed-Time Optimal Formation Synchronous Tracking Control of UAV-Ugvs Based on Reinforcement Learning

Abstract

This paper explores the issue of prescribed-time optimal formation synchronous tracking control of nonlinear unmanned aerial vehicle (UAV) and unmanned ground vehicle (UGV) systems. Based on a novel constructed second-order nonlinear UAV-UGV swarm case, the reinforcement learning (RL) method is introduced to obtain the optimal control scheme, and a gradient descent method with a simple positive function for the Hamilton-Jacobi-Bellman (HJB) equation is improved to establish the adaptive actor and critic networks and solve the iterate adaptive laws, which allows adaptive parameters to be trained more thoroughly. Integrating the prescribed-time constraints of formation tasks, the incorporation of traditional prescribed-time functions can result in structural modifications within the RL framework and state coupling, thereby increasing the complexity of control strategy design. Hence, the prescribed-time functions are designed in the critic and actor networks, which address the state coupling and optimize the acquisition of adaptive parameters under the gradient descent method. Then, by employing the aforementioned methods, an optimal synchronous control scheme is proposed to address nonlinear UAV-UGV time-varying formation tracking at a settling time, and a pivotal scaling technique is used for formation stability analysis. Finally, simulation and experiment results are carried out to demonstrate the efficacy of the proposed approach.