Robot Behavior During Human-Robot Object Handover

When we hand an object to another human we let go only when we know the recipient has achieved a secure grasp. How do we "know" that the recipient has a secure grasp on the object? Humans rely on touch sensing to monitor and mediate this "handover" process. A subtle aspect of this process is that we use "active perception" to detect the quality of the recipient's grasp. By imposing small movements on the object being passed and observing the changes in (touch) sensory information, we can detect when the recipient has achieved a sufficiently secure grasp, such that we can safely let go of the object. As robots begin to assist humans with physical tasks, such as surgery, warehousing, household aid, etc., robot-to-human object handovers will become common. For these handovers to be safe and efficient the robot must be able to perceive the quality of the human recipient's grasp on the object. This project will develop active (haptic) perception methods that will enable a robot to measure the "degree-of-constraint" imposed by the recipient on the object and thus detect when it is safe to pass the object.