Design of action policies of a robot to collaborate with human partners while being in physical contact with the human (From A Sensorimotor Reinforcement Learning Framework for Physical Human-Robot Interaction A. Ghadirzadeh, J. Butepage, A. Maki, D. Kragic and M. Björkman, https://arxiv.org/pdf/1607.07939.pdf)
In the near future robots will enter into our daily lives and we need to make them more socially compatible with human. In order to safely and meaningfully interact with humans, robots must develop an advanced real-world social intelligence and be transparent with respect to the causes and reasons for their actions to human users, i.e. they must be able to communicate in terms that humans find intuitive and understandable. One of the research ideas we follow in this area is an action-oriented approach to social cognition in biological and artificial agents. We are inspired by a view that assumes that even complex modes of social interaction are grounded in basic sensorimotor patterns enabling the dynamic coupling of agents. Such sensorimotor patterns, or contingencies are known to be highly relevant in cognition. One of the projects studying this is Socializing Sensorimotor Contingencies (
socsmcs.eu ) that follows a key hypothesis that learning and mastery of action-effect contingencies are also critical to enable effective coupling of agents in social contexts.
We are currently involved in the followingprojects:
EU H2020 socSMCs
EU FP7 TRADR
For more information, please contact involved faculty members: