Neural-FxSMC: A Robust Adaptive Neural Fixed-Time Sliding Mode Control for Quadrotors with Unknown Uncertainties

This paper presents Neural-FxSMC, a robust and precise control scheme for quadrotors to counter unknown dy- namics, uncertainties, and external disturbances. Neural-FxSMC, (i) addresses fixed-time convergence of the tracking error, control singularity, and chattering issues simultaneously, which is not possible with the existing Fixed time Sliding Mode Control (FxSMC), and (ii) relaxes the a priori bound assumption over the uncertainties that are often considered as a constant or a state- dependent upper bound. The fixed-time convergence of tracking error is guaranteed by establishing fixed-time convergence of the Non-singular Fast Terminal Sliding Surface (NFTSS), contrary to the existing works where the NFTSS convergence depends on initial conditions. The Chattering is suppressed via Radial Basis Function Network (RBFN) based uncertainties estimation. Finally, using the Lyapunov theory, we prove the fixed-time convergence and boundedness of Neural-FxSMC weights. We comprehensively evaluate Neural-FxSMC in challenging scenarios such as unknown payload and turbulent wind. Our Neural- FxSMC, apart from handling unknown dynamics and uncer- tainties, also offers direct gravity compensation without using quadrotor mass and gravity.