Human-Robot Interaction With Robust Prediction of Movement Intention Surpasses Manual Control

Abstract

Physical human-robot interaction (pHRI) enables a user to interact with a physical robotic device to advance beyond the current capabilities of high-payload and high-precision industrial robots. This paradigm opens up novel applications where a the cognitive capability of a user is combined with the precision and strength of robots. Yet, current pHRI interfaces suffer from low take-up and a high cognitive burden for the user. We propose a novel framework that robustly and efficiently assists users by reacting proactively to their commands. The key insight is to include context- and user-awareness in the controller, improving decision-making on how to assist the user. Context-awareness is achieved by inferring the candidate objects to be grasped in a task or scene and automatically computing plans for reaching them. User-awareness is implemented by facilitating the motion toward the most likely object that the user wants to grasp, as well as dynamically recovering from incorrect predictions. Experimental results in a virtual environment of two degrees of freedom control show the capability of this approach to outperform manual control. By robustly predicting user intention, the proposed controller allows subjects to achieve superhuman performance in terms of accuracy and, thereby, usability.