24 November 2006 (week 8)
One of the great challenges in bringing robots out of laboratories into real-world domestic and public environments is to address the fact that humans and robots have different modalities, and exploit this knowledge for successfully operate in the same space.
This talk will present current work by the Adaptive Systems research group at the University of Hertfordshire on social interaction and social learning between humans and robots, in shared environments. Imitation is a powerful learning tool when a number of agents interact in a social context. Robots capable of imitation (or other, simpler, forms of social learning) would allow humans to interact with them and transfer knowledge in a more adaptive and "natural" (for the humans) way.
Here, for our purposes, the agents (humans and robots) have dissimilar bodies and affordances, making it necessary to acknowledge and address the 'correspondence problem' for each such interaction. We discuss the social aspects of humans teaching robots and propose some initial requirements for the contextual interpretation of body postures and human activities (within the human-robot teaching context).