This paper addresses the challenging concept of Assist-As-Needed control of exoskeleton robots. The proposed controller boosts the voluntary participation of the patient by providing assistance according to the ability of the wearer in performing the assigned task. A novel strength index is presented that combines interaction force and position-tracking error into a single quantity to estimate the physical strength of the wearer during the therapy. The estimated strength is used to adjust the boundaries of a virtual tunnel around the desired trajectory, defined to assume a degree of freedom for the wearer’s motions and to compensate for asymmetric gait patterns. The required assistance is then defined by an adaptive impedance controller according to the distance of the tracking error from the tunnel boundaries. To ensure that the assistance is accurately supplied to the patient, an adaptive torque controller is integrated into the control loop. The adaptive torque controller uses a generalized fuzzy hyperbolic model to compensate for the inherent impedance of the exoskeleton. Simulation results on a hemiplegic model show that the proposed index can estimate the wearer’s strength properly and the proposed assist-as-needed controller can reduce the tracking error. The performance of the proposed method is also evaluated experimentally on a healthy subject wearing a hip exoskeleton. The results verify that the proposed method can be used in a variety of therapeutic applications where it is important to track the desired trajectory with minimum interventions.

Keywords