Adaptive dynamic programming-based optimal pursuit–evasion control for quadrotor unmanned aerial vehicles with obstacle avoidance

This article investigates the challenging problem of optimized intelligent pursuit–evasion control for quadrotor unmanned aerial vehicles with obstacle avoidance. Firstly, a novel penalty function is developed to achieve obstacle avoidance in the designed cost function. Subsequently, a critic-only structure is employed to substitute the conventional actor-critic structure for addressing the approximate solutions of Hamilton–Jacobi–Isaacs equations. Meanwhile, two new critic neural networks are constructed to learn the optimal cost functions and control policies for the pursuit–evasion control systems. More specifically, the developed weight update laws not only enable the critic neural network weights to be updated online, but also alleviate the persistent excitation condition with a simple structure. In particular, the obstacle avoidance function is incorporated into the construction of the approximate cost function learned by the proposed critic neural networks. In addition, the proposed pursuit–evasion control systems can be guaranteed to achieve uniformly ultimately bounded stability by utilizing Lyapunov methodology. Finally, the effectiveness of the developed pursuit–evasion control scheme is fully illustrated through two simulation cases.