A study on the use of tactile instructions for developing robot's motions

Developing motions for humanoid robots is a time consuming task. However, we note that sport or dance instructors can easily adjust their students' postures by simple touches. This suggests the possibility of exploiting touch for motion development, and allows us to propose a methodology based on this concept. This requires defining how the robot should interpret user's touches. We propose a supervised learning approach for coping this issue, and verify its feasibility experimentally. We then study the data collected by the algorithm, and show that the system is usable both for motion development and as a tool for studying human robot tactile communication. In particular, considerations on the sparsity that characterize the whole process are presented, and policies for an efficient interpretation of tactile instructions are drawn.