Human-in-the-Loop Prosthetic Robot Hand Control Using Particle Filters for Grasp Selection

摘要

Human intent inference is a key factor in shared autonomy. The control of robotic prosthetic hands with inherent human-in-the-loop operations is not an exception. State-of-the-art methods use electromyography (EMG) signals to infer grasp types. But this approach is limited by important robustness issues which renders its use limited outside labs. In this paper, we use a particle filtering based technique to infer user intent based on the trajectories of the user's hand. Our methodology generates an estimation of the expected time-to-arrive, i.e. time left until reaching to the object, which is an essential variable in successful grasping of objects. The approach is integrated using a probabilistic framework which can incorporate information from all available information sources. We have tested our algorithm with 80 reaching trajectories collected from 5 individuals reaching to one of four objects at a time. The results demonstrate a classification accuracy of 87.5% for the entire reaching period.