Integrated Neural Fault-Tolerant Control Using LPV Sliding Mode Observer for Quadrotors

This paper proposes an integrated linear parameter varying sensor fault-tolerant control scheme that combines a parameter-varying sliding mode observer with a baseline neural network controller. The scheme aims to address the degradation of fault estimation performance and nominal closed-loop performance caused by sensor faults/failures despite external disturbances. This is achieved by simultaneously parameterizing the observer gains and perturbations of weighting matrices in the neural network controller. To ensure numerical tractability, the scheme transforms a nonconvex parameter-varying bilinear matrix inequality problem into a linear matrix inequality formulation. By solving a finite number of linear matrix inequalities in one step, robustness performance is guaranteed in an <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\mathcal {L}_{2}$</tex-math></inline-formula> sense. Furthermore, stability analysis of the scheme is rigorously conducted using a Lur'e-Postnikov-like Lyapunov function to enhance conservatism when dealing with sector-bounded and slope-restricted nonlinear activation functions. A quadrotor position control example involving direct neural network control under scenarios of sensor bias and total loss of sensor measurements illustrates the effectiveness of the scheme.