Adaptive Command Filter Backstepping Sliding Mode Control for Variable Stiffness Exoskeleton With Input Saturation Constraint

The variable stiffness exoskeleton exhibits excellent performance in terms of high safety, compliance, task adaptability, and robustness during physical human-robot interaction (pHRI). However, the position control of the variable stiffness exoskeleton is challenging due to the problems of real-time variable stiffness, model uncertainties, unknown model parameters, disturbances, actuator failure, and input saturation. In this article, an adaptive command filter backstepping sliding mode control based on disturbance compensation and input saturation constraint is proposed. First, in order to solve the problem of input saturation, a function approximation method is used to transform the system. Second, a method combining disturbance observer (DOB) and neural network (NN) is proposed to estimate and compensate the lumped disturbances of the system. The estimator can compensate for lumped disturbances in the link subsystem and also compensate for actuator fault in the motor subsystem. Third, the inherent “complexity explosion” and filtering error problems in backstepping are solved by command filtering and filter compensation method. At the same time, the stability of the proposed method is proved by Lyapunov method. Finally, simulations and experiments are conducted in three typical application cases to verify the effectiveness of the proposed method.