Neural-network-based sliding-mode control for multiple rigid-body attitude tracking with inertial information completely unknown

This paper addresses the multiple rigid bodies attitude tracking control problem in presence of inertial information completely unknown. In order to attenuate the effect of unknown parameters, the neural networks technology is employed to approximate the unknown nonlinear terms derived from the controller design procedure. Based on lumped tracking errors between neighbors, a novel sliding-mode control protocol is proposed and an approximate expression of inverse matrix is also applied to avoid possible singular phenomenon. The Lyapunov theory is applied to guarantee that all signals in the closed-loop system are uniformly ultimately bounded and that all rigid-body attitude synchronize to the desired trajectory with bounded residual errors. Compared with prior works, the dynamics of each rigid-body discussed here is more general and does not require the assumption linearity in the unknown parameters or the matching condition. Moreover, the bounded residual errors can be reduced as small as desired by selected suitable parameters. Simulation results demonstrate the effectiveness of the proposed method.