Improving Human-Robot Interaction in Upper-Limb Rehabilitation Exoskeletons through Human-Like Path Generation and Patient-Cooperative Control

Abstract

Exoskeleton-based therapy is a growing area and can provide benefits over conventional manual therapy due to the robots’ ability in providing high intensity and long therapy sessions, accurate measurements, and precise control over individual joints. Despite the recent progress in the development of upper-limb rehabilitation exoskeletons, their employment in rehabilitation settings is limited. Among various reasons, one main limiting factor is the problem of human-robot interaction in these systems, which hasn’t been studied thoroughly. From one perspective, improving human-robot interaction requires developing systems that conform to the user. The alignment of the robotic and human joint centers is the requirement for achieving kinematic compliance and ergonomic design. This problem is, however, challenging due to the complexity of the biological joints such as the shoulder. Improved interaction can also be achieved by using exoskeleton motion generation algorithms which can produce motions similar to the natural motions of the users. Additionally, developing control algorithms that better resemble the supervisory role of the therapist during manual rehabilitation, can provide a more compliant therapy procedure and will further expand the usability of exoskeletons in the rehabilitation settings. This work addresses the three mentioned areas of improvement by offering unique solutions in each area. First, to develop an ergonomic upper-limb exoskeleton that can correspond to the functionality of the human arm and move in accordance with the upper-limb joints, a kinematic design including a new inner shoulder design is introduced based on the systematic approaches of product development. The developed kinematic design is examined through kinematic and experimental analysis on the design. The novel design of this new upper-limb exoskeleton, called CLEVERarm, allows for accurate alignment in the shoulder joint by moving the device joints in harmony with the human body. Additionally, a new control framework is developed which enables automated and patient-cooperative rehabilitation. For this purpose, a new computational-based reference path generation method for upper-limb exoskeletons is introduced which generates human-like reference motions for the exoskeleton. The proposed method generates human-like motions in the configuration space of the human arm and transfers the generated motions to the configuration space of the exoskeleton. The accuracy of the introduced model is verified through experimental and simulation results. Additionally, a new performance-based compliant controller is developed based on variable admittance control. The developed controller updates the prescribed reference path to be followed, based on the user exerted forces, to comply with the user’s movement. However, deviations from the reference path are limited and incorrect posture is not permitted. The performance-based and assist-as-needed nature of the developed controller encourages patient contribution. Experimental verification results show the capability of the developed algorithm to be employed in upper-limb exoskeletons for the rehabilitation of patients with various levels of motor impairment.