Measuring intent in human-robot cooperative manipulation

To effectively interact with people in a physically assistive role, robots will need to be able to cooperatively manipulate objects with a human partner. For example, it can be very difficult for an individual to manipulate a long or heavy object. An assistant can help to share the load, and improve the maneuverability of the object. Each partner can communicate objectives (e.g., move around an obstacle or put the object down) via non-verbal cues (e.g., moving the end of the object in a particular direction, changing speed, or tugging). Herein, non-verbal communication in a human-robot coordinated manipulation task is addressed using a small articulated robot arm equipped with a 6-axis wrist mounted force/torque sensor and joint angle encoders. The robot controller uses a Jacobian Transpose velocity PD control scheme with gravity compensation. To aid collaborative manipulation we implement a uniform impedance controller at the robot end-effector with an attractive force to a virtual path in the style of a cobot. Unlike a cobot, this path is recomputed online as a function of user input. In our present research, we utilize force/torque sensor measurements to identify intentional user communications specifying a change in the task direction. We consider the impact of path recomputation and the resulting robot haptic feedback on user physiological response.