Development of an exoskeleton robot for upper-limb rehabilitation

Abstract

To assist or rehabilitate individuals with impaired upper-limb function, we have developed an upper-limb exoskeleton robot, the ETS-MARSE (motion assistive robotic-exoskeleton for superior extremity). The MARSE is comprised of a shoulder motion support part, an elbow and forearm motion support part, and a wrist motion support part. It is designed to be worn on the lateral side of the upper limb in order to provide naturalistic movements of the shoulder (i.e., vertical and horizontal flexion/extension, and internal/external rotation), elbow (i.e., flexion/extension), forearm (i.e., pronation/supination), and wrist joint (i.e., radial/ulnar deviation, and flexion/extension). This thesis focuses on the modeling, design (mechanical and electrical components), development, and control of the developed MARSE.
\n
\nThe proposed MARSE was modeled based on the upper-limb biomechanics; it has a relatively low weight, an excellent power/weight ratio, can be easily fitted or removed, and is able to effectively compensate for gravity. Moreover, to avoid complex cable routing that could be found in many exoskeleton systems, a novel power transmission mechanism was introduced for assisting shoulder joint internal/external rotation and for forearm pronation/supination. The exoskeleton was designed for use by typical adults. However, provisions are included for link length adjustments to accommodate a wide range of users. The entire exoskeleton arm was fabricated primarily in aluminum except the high stress joint sections which were fabricated in mild steel to give the exoskeleton structure a relatively light weight. Brushless DC motors (incorporated with Harmonic Drives) were used to actuate the developed MARSE.
\n
\nThe kinematic model of the MARSE was developed based on modified Denavit-Hartenberg notations. In dynamic modeling and control, robot parameters such as robot arm link lengths, upper-limb masses, and inertia, are estimated according to the upper limb properties of a typical adult. Though the exoskeleton was developed with the goal of providing different forms of rehab therapy (namely passive arm movements, active-assisted therapy, and resistive therapy), this research concentrated only on passive form of rehabilitation.
\n
\nPassive arm movements and exercises are usually performed slowly compared to the natural speed of arm movement. Therefore, to control the developed MARSE, a computationally inexpensive a PID controller and a PID-based compliance controller were primarily employed. Further, realizing the dynamic modeling of human arm movement which is nonlinear in nature, a nonlinear computed torque control (CTC) and a modified sliding mode exponential reaching law (mSMERL) techniques were employed to control the MARSE. Note that to improve transient tracking performance and to reduce chattering, this thesis proposed the mSMERL, a novel nonlinear control strategy that combined the concept of boundary layer technique and the exponential reaching law. The control architecture was implemented on a field-programmable gate array (FPGA) in conjunction with a RT-PC.
\n
\nIn experiments, typical rehabilitation exercises for single and multi joint movements (e.g., reaching) were performed. Experiments were carried out with healthy human subjects where trajectories (i.e., pre-programmed trajectories recommended by therapist/clinician) tracking the form of passive rehabilitation exercises were carried out.
\n
\nThis thesis also focused on the development of a 7DoFs upper-limb prototype (lower scaled) ‘master exoskeleton arm’ (mExoArm). Furthermore, experiments were carried out with the mExoArm where subjects (robot users) operate the mExoArm (like a joystick) to maneuver the MARSE to provide passive rehabilitation.
\n
\nExperimental results show that the developed MARSE can effectively perform passive rehabilitation exercises for shoulder, elbow and wrist joint movements. Using mExoArm offe