Magni Dynamics: A Vision-Based Kinematic And Dynamic Upper-Limb Model For Intelligent Robotic Rehabilitation

Abstract

This paper presents a home-based robot-rehabilitation<br> instrument, called ”MAGNI Dynamics”, that utilized a vision-based<br> kinematic/dynamic module and an adaptive haptic feedback<br> controller. The system is expected to provide personalized<br> rehabilitation by adjusting its resistive and supportive behavior<br> according to a fuzzy intelligence controller that acts as an inference<br> system, which correlates the user’s performance to different stiffness<br> factors. The vision module uses the Kinect’s skeletal tracking to<br> monitor the user’s effort in an unobtrusive and safe way, by estimating<br> the torque that affects the user’s arm. The system’s torque estimations<br> are justified by capturing electromyographic data from primitive<br> hand motions (Shoulder Abduction and Shoulder Forward Flexion).<br> Moreover, we present and analyze how the Barrett WAM generates<br> a force-field with a haptic controller to support or challenge the<br> users. Experiments show that by shifting the proportional value,<br> that corresponds to different stiffness factors of the haptic path, can<br> potentially help the user to improve his/her motor skills. Finally,<br> potential areas for future research are discussed, that address how<br> a rehabilitation robotic framework may include multisensing data, to<br> improve the user’s recovery process.