Goal-Directed Tactile Exploration for Body Model Learning Through Self-Touch on a Humanoid Robot

An early integration of tactile sensing into motor coordination is the norm in animals, but still a challenge for robots. Tactile exploration through touches on body gives rise to first models and bootstraps further development such as reaching competence. Reaching one’s own requires connections space only. Still, problems high dimensionality redundancy persist. Through an embodied computational model learning self-touch simulated humanoid robot with artificial sensitive skin, we demonstrate that this task can be achieved (i) effectively (ii) efficiently at scale by employing frameworks internal reaching: intrinsic motivation goal babbling. We relate our results infant studies spontaneous well vibrotactile targets body. analyze configurations one followed weekly between 4 18 months age derive requirements model: accounting (iii) continuous rather than sporadic touch (iv) consistent resolution. Results show general success domain, also point out limitations achieving fully touch.