Recognizing Collaboration Intent to Control Physical Human-Robot Interaction

In recent years, there has been intense interest in collaborative robots, both for industry and household applications. While significant progress has been made, physical human-robot interaction is still presenting a challenging problem that has not been satisfactorily solved. When a human is interacting with another human, the forces they exchange represent a communication channel and a continuous stream of information flows between them. When a human is interacting with a robot, the forces applied by the robot are interpreted by the human that in turn reacts to them; obviously, people are expecting the robot to also react to the forces they are applying. In this research, we identify different types of collaboration during collaborative manipulation and use this information to better control human-robot interaction. We propose a new metric for the identification of the cooperation intent and study how to best compute the interaction force, on which our metric is based, in a real-time application. We also propose a control framework that uses a set of robot controllers that are selected using the identified collaboration intent to control the robot during collaborative tasks. Finally, we present our preliminary experiments with the Baxter robot. The experiments have been performed in order to understand the precision, repeatability and safety of the robot using different control approaches. These experiments informed the proposed controllers and are the key for their future implementation on the Baxter robot.