Dynamic Movement Primitives and Adaptive Nonsingular Fast Terminal Sliding Mode Control for Stable Trajectory Generalization and Tracking of Lower‐Limb Exoskeletons

Abstract

Personalized gait generation is required in exoskeleton‐assisted training because of intersubject differences and changes across rehabilitation stages. This article proposes a trajectory generalization and tracking method that integrates dynamic movement primitives (DMPs) with adaptive nonsingular fast terminal sliding mode control (ANFTSMC). First, reference trajectories are constructed using a five‐point segmented gait‐planning strategy. DMPs are then introduced for trajectory generalization, and the DMP weights are dynamically optimized according to the zero moment point (ZMP) and a stability‐margin criterion. Second, to address tracking errors caused by human–exoskeleton interaction forces and unmodeled factors, the identified joint dynamic modeling terms are used as feedforward compensation. A nonlinear disturbance observer (NDO) is employed to estimate the lumped disturbance online, and disturbance compensation and trajectory tracking are achieved through the ANFTSMC control law. Finally, the effectiveness of the controller is cross‐validated through simulations and wearable experiments: the simulations characterize gait motion stability, whereas the wearable experiments evaluate joint tracking errors and human–exoskeleton interaction torque responses under controller action. The results show that ANFTSMC‐NDO outperforms the other methods in tracking accuracy and response speed. The wearable experiments also yield lower human–exoskeleton interaction torque responses, verifying the tracking performance, compensation capability for unmodeled effects, and experimental feasibility of the generated generalized gait.