Data-driven safety control with velocity and input force constraints for a rehabilitation walker considering human-robot training environment

This study investigates a data-driven controller for a rehabilitation walker with constraints in relation to velocity and input force. By decomposing the generalised input force and coefficient matrix in the dynamic model, a mathematical description of the human-robot training environment was constructed. A new equivalent data-driven model of the walker considering the human-robot training environment was established by using pseudo-Jacobian matrix. The data-driven estimation method for variations in the human-robot training environment was proposed by designing an extended state observer, and the stability of the estimation error system was analysed. In addition, the desired motion velocity of the walker was designed. This approach has the advantage that the safety of the specified velocity can be guaranteed. We proposed the data-driven tracking controller that simultaneously constrains the actual velocity and input force while stabilising the tracking error system. We also illustrated the effectiveness of the proposed data-driven algorithm, and whose effectiveness was further verified by simulation-based comparative analysis.