To achieve safe close-proximity human-robot interaction , particularly for physically assitive tasks, robot motion planning needs to recognize and adapt to the behaviour of humans in the long-term. Generally, motion prediction with probabilistic confidence-awareness models can be used to reason and predict with a degree of confidence the next movement of human in task. As such in order to ensure safety in robot trajectory planning, it is important to learn the variance in the human movement. The variance in the human skill to interact with robots in close proximity can drastically be altered if external distractions or cognitive overloading occur during such interaction. Even if this can be a one-off event, it is critical to evaluate the alteration in the movement to ensure long-term safety over the long-term. For robust long-term human movement prediction and adaption, the overall safety properties of any planner requires collision-free, good approximation and coupling of the robot's movement with that of the human when these type of mistakes occur. In the context of close-proximity human-robot collaboration, the confidence-aware models need to have a measure of cooperation by observing changes in human-robot trajectory comparison because in such interaction collision avoidance is not a requirement of the interaction, but the synchronous movement is. We provide a collaborative measure that represents coupling between the robot movement and the human movement. This measure can provide an additional metric for ensuring overall safety in trajectory planning. We demonstrate this cooperative measure in an assistive dressing scenario, where controlled experiments were performed to study the differences in coupling during learning of the robot-initiated dressing skill by the human, and then introducing cognitive loading and external distractions. This cooperative measure helps to detect the lack of collaboration caused by the one-off events caused by a lack of cooperation rather than just accounting for just the variance in the movement of a learned skill.
Camilleri, A., Hong, J., Dogramadzi, S., & Caleb-Solly, P. (2020, May). Towards establishing a 'cooperation' measure for coupled movement in close-proximity human-robot interaction. Presented at ICRA 2020