RBF Network-Based Adaptive Control for Humanoid Gait Data

In order to solve the trade-off between model uncertainty compensation and vibration suppression for sliding mode control in lower limb rehabilitation robots, an adaptive sliding mode control strategy based on radial basis function (RBF) neural network is proposed in this paper. By combining the strong robustness of the sliding mode control with the nonlinear approximation ability of the RBF network, the proposed method ensures high-precision trajectory tracking while significantly reducing the vibration amplitude. Firstly, an exoskeleton-human body coupling dynamic model is established, and an integral sliding mode surface containing the rehabilitation trajectory tracking error is designed. Then, an RBF network was employed to approximate the unmodelled dynamic characteristics and external perturbations in real time. The adaptive weight updating law and switching gain adaptation law based on the neural network effectively solved the challenges of accurate trajectory tracking and appropriate human-machine interaction force in rehabilitation training, balanced the motion accuracy and wearing comfort, and was particularly suitable for rehabilitation training of patients with neurological injuries.