Learning sensorimotor navigation using synchrony-based partner selection

Future robots are supposed to become our partners and share the environments where we live in our daily life. Considering the fact that they will have to co-exist with "nonexpert" people (elders, impaired people, children, etc.), we must rethink the way we design human/robot interactions. In this paper, we take a radical simplification route taking advantage from recent discoveries in low-level human interactions and dynamical motor control. Indeed, we argue for the need to take the dynamics of the interactions into account. Therefore, we propose a bio-inspired neuronal architecture to mimics adult/infant interactions that: (1) are initiated thanks to synchrony-based partner selection, (2) are maintained and re-engaged thanks to partner recognition and focus of attention, and (3) allow for learning sensorimotor navigation based on place/action associations. Our experiment shows good results for the learning of a navigation area and proves that this approach is promising for more complex tasks and interactions.