Adaptive Admittance Control of an Upper Extremity Rehabilitation Robot With Neural-Network-Based Disturbance Observer

Robot-assisted rehabilitation therapy has become an important technology applied to recover the motor functions of disabled individuals. In the present paper, an adaptive admittance control strategy combined with a neural-network-based disturbance observer (AACNDO) is developed for a therapeutic robot to provide upper extremity movement assistance. Firstly, a comprehensive overview of the robot hardware and real-time control system is introduced. Then, the dynamics-based adaptive admittance controller is designed to improve human-robot interaction compliance and induce the active participation of the patients during rehabilitation training. A disturbance observer with a radial basis function network is designed to guarantee the control performance with external uncertainties and dynamics error. Besides, an adaption law is integrated into the admittance model to adjust the interaction compliance in different working areas based on the motion intention and recovery phase of the patient. Further experimental investigations, including sinusoidal trajectory tracking experiments, circular trajectory tracking experiments with admittance adjustment, and intention-based resistive training experiments, are conducted by three volunteers. Finally, the experimental results validate the feasibility and effectiveness of the rehabilitation robot and AACNDO scheme in providing patient-passive and patient-cooperative rehabilitation training.