"Exoskeleton Controller Data"

"Safe and compliant human-robot interaction in lower-limb exoskeletons encounters dynamic parameter uncer\u00adtainties and nonlinear friction, degrading position tracking accuracy and leading to excessively high interaction impedance. In order to solve this problem, this paper proposes a state-con\u00adstrained observer based adaptive neural admittance control strategy. In particular, we use an outer-loop admittance model that maps interaction forces to desired trajectories, such that the desired compliant behavior is achieved. To achieve high-fidelity transparency, a composite inner-loop controller is established. Specifically, by leveraging the velocity tracking error as a sensitive indicator of interaction lag, a Nonlinear Disturbance Observer (NDO) acts as a feedforward compensator to actively 'mask' the exoskeleton's inherent friction and inertia. This is further augmented by a Radial Basis Function Neural Network (RBFNN) to handle unstructured high-frequency dynamicsMoreover, the operational safety is theoretically guaranteed by the inclusion of a Barrier Lyapunov Function (BLF), which strictly enforces the position and velocity constraints. Both simulations and experiments show us that compared with the standard backstepping control, the proposed approach lowers the RMSE and MAX on position tracking by 48.7% and 35.88%, respectively. Most significantly, the method is quite effective in reducing system impedance, achieving a 49.52% reduction in human-robot interaction torque, thus ensuring enhanced human interaction compliance and safety."