An Assist-as-Needed Velocity Field Control Scheme for Rehabilitation Robots

This paper addresses the problem of assist-as-needed (AAN) control for rehabilitation robots. To achieve a motion which is not explicitly a function of time, the velocity field control is considered in this paper. The proposed new controller consists of a proportional-like feedback term and a neural network (NN) term, where the later is exploited to compensate for the dynamic uncertainties of the system. The AAN property is facilitated by means of a dead-zone function in the feedback control term and a forgetting factor in the adaptation law of NN component. The designed controller guarantees the stability of the system with a bounded control command. The performance of the proposed AAN scheme is validated through the simulation and experiment conducted on a lower-limb exoskeleton.