Recurrent Neural Network Based Sliding Mode Control for an Uncertain Tilting Quadrotor UAV

ABSTRACT In this paper, a new recurrent neural network‐based sliding mode control (RNN‐based SMC) strategy for performing the desired position and attitude tracking control of an uncertain tilting quadrotor unmanned aerial vehicle (TQUAV) with various situations is proposed. The key research objective is to pursue the adaptive adjustment mechanism of the sliding mode manifold parameters, which are usually and directly used as constants in the existing literature. To achieve this objective, the constructed manifold parameters and the lumped disturbances are approximated by utilizing RNN, the corresponding approximation errors derived from RNN are estimated by employing an adaptive control method, and the compensatory controllers are further designed to guarantee the convergence of all position and attitude tracking errors. The RNN‐based SMC strategy has the capabilities of adjusting the controller parameters in real‐time, guaranteeing the controller continuity, also tracking the desired trajectories of uncertain TQUAV robustly and adaptively. Finally, the effectiveness of the RNN‐based SMC strategy is fully verified through theory and comparative simulation results.