Dynamic Event-Triggered Robust Optimal Attitude Control of QUAV Using Reinforcement Learning

In this article, a dynamic event-triggered robust optimal attitude tracking control problem for a quadrotor unmanned aerial vehicle in an uncertain environment is investigated. First, an augmented system consisting of tracking error signal and reference signal is developed to transform the tracking problem into a stabilization problem. Then, in order to handle the random disturbances in the design of the optimal controller, a new Hamilton–Jacobi–Bellman equation is proposed. Subsequently, the dynamic event-triggered mechanism is designed to reduce the communication pressure, and an event-based critic-only reinforcement learning algorithm is proposed to implement the optimal controller design. Remarkably, by combining concurrent learning technique and gradient descent algorithm, the adaptive weight update law is derived to tune the critic neural network, thus erasing the demand on the persistent excitation condition. After that, we demonstrate that the closed-loop system is semiglobally uniformly ultimately bounded in mean square and prove the Zeno-free behavior. Finally, the simulation results are given to show the effectiveness of our control strategy.