Reinforcement learning based - sliding mode control for trajectory tracking of quadrotor unmanned aerial vehicles under disturbances

In this article, a reinforcement learning (RL)-based sliding mode control (SMC) is proposed for trajectory tracking of a quadrotor unmanned aerial vehicle (QUAV) under external disturbances. First, an actor-critic RL framework sliding mode control is provided to tackle the optimal control problem without external disturbances. Secondly, the simulation in an environment with disturbances is carried out to show the robustness of the proposed controller. Theoretical analysis shows that the position and attitude tracking errors converge to a preset region, and the weight estimation errors of the actor-critic networks are uniformly ultimately bounded (UUB). Finally, a comparison of the numerical simulations between the proposed controller and traditional sliding mode controller and the Backstepping (BSP) technique is provided to indicate the advantages and improved performance of the RL-based SMC.