A Robot-Aided Rehabilitation Platform for Occupational Therapy with Real Objects

Robot-aided rehabilitation enables to assist the patient in executing task-oriented exercises and typically takes advantage of virtual environments in which motor tasks are performed. However, this approach introduces a mismatch between the visual and the proprioceptive stimuli the patient perceives. The use of real tools to perform rehabilitation tasks could overcome this drawback. Nevertheless, several issues arise, such as the estimation of the pose of the real objects in the workspace and the planning of the trajectories the robot has to execute to guide the patient’s limb to reach the objects. In this paper, a robot-aided upper-limb rehabilitation system able to recognize the objects the patient has to interact with, by means of an RGB-D camera, and to dynamically plan the robot trajectories is proposed. The computational burden and the performance of the proposed system, and in particular of the Pose Estimation Pipeline and of the robot Motion Planner, are evaluated. The obtained results in terms of pose estimation and reaching errors pave the way to the application of the proposed system in a real rehabilitation scenario.