Programming robot motions by touching

Small humanoid robots are becoming a more and more pop- ular device for entertainment. While the literature shows many advanced techniques for the development of humanoid robot motions, for example motion retargeting, most of the commercially available products only provide a low level in- terface where each motion is described in terms of keyframes for which each joint angle is defined. In this paper we suggest to employ tactile interaction as an intuitive way for users to communicate with robots. The interface we developed serves two purposes: it allows the user to develop robot motions in a very natural way, and it allows us to collect data use- ful for studying the characteristics of touching as a means of communication. A supervised learning algorithm suited to extract the meaning of touch patterns is presented, and preliminary experiments that validate the feasibility of the approach are presented.