Design, implementation and BCI-based control of a series elastic mobile robot for home-based physical rehabilitation

Official URL: https://risc01.sabanciuniv.edu/record=b1558924_(Table of contents)

We present the design, control and human-machine interface of AssistOn- Mobile, a series elastic holonomic mobile platform aimed to administer therapeutic table-top exercises to patients who have suffered injuries that affect the function of their upper extremities. AssistOn-Mobile is designed as a multi-DoF series elastic actuator based on a holonomic mobile platform consisting of four actuated Mecanum wheels. Thanks to its mobile base, it is a compact and portable device that can cover whole human workspace for planar reaching exercises. Even though the mobile platform is not passively backdriveable due to the Mecanum wheels, a low-cost, compliant mechanism based, multi-DoF series elastic element is introduced to transform the holonomic mobile platform into a series elastic actuator and to ensure actively backdriveability of the overall system. Implementation of series elastic actuation results in many advantages, such as eliminating the need for costly force sensors, improving the performance of force control, and increasing the impact resistance and robustness of the overall system. For the control of AssistOn-Mobile, in addition to admittance controllers used for active backdriveability, passive velocity eld control (PVFC) is implemented for contour following tasks. PVFC minimizes the contour error by decoupling the task (path tracking) and the timing of the task, while also ensuring the coupled stability of the human-in-the-loop system by rendering the system passive with respect to externally applied forces. Furthermore, since AssistOn-Mobile is an end-effector type device, patients' shoulder movements are continually tracked utilizing a Kinect (RGBD) sensor such that compensatory movements of the patients (e.g, leaning) are limited by providing online feedback to the patients (for instance, by modulating the speed of the contour tracking task). With these controllers in place, AssistOn-Mobile becomes a highly backdriveable, force-controlled robotic interface that can provide required amount of assistance/ resistance to patients, while performing omni-directional movements on a plane. To enable patients with severe disabilities (e.g., spinal cord injury patients with no residual movements on their affected limb) to interact with AssistOn- Mobile and to provide assist-as-needed rehabilitation protocols to such patients, we introduce a systematic approach for online modi cation of robot assisted rehabilitation exercises by continuously monitoring intention levels of patients utilizing an electroencephalogram (EEG) based Brain-Computer Interface (BCI). In one implementation, we utilize posterior probabilities extracted from an LDA classiffier as the continuous-valued outputs to PFVC. This way, PVFC modulates the speed of contour following tasks with respect to intention levels of motor imagery. The efficacy of our proposed robotic BCI framework with online modification of task speed is investigated by a set of human subject experiments with healthy volunteers. In particular, our approach is compared with the existing BCI-based virtual reality and robot-assisted rehabilitation techniques. Within this experiment, we have collected statistically significant evidence of the beneficial effect of the haptic feedback during the mental imagery of subjects. Results also indicate that using BCI continuously rather than to initialize the movement only may be preferable to ensure active participation of patients throughout the therapy. Finally, using the proposed BCI-based rehabilitation protocol shows no statistically signi cant difference in terms of mental imagery activity, compared to the rehabilitation protocol where the subjects are actively performing the real movement.