Chris Adams



social robots:

learning to communicate


learning to communicate


Contact: Aude Billard
Last updated: Fri Mar 3 12:04:20 2000


Most living societies communicate, from insects to humans. In low level species, communication abilities seem often built-in. However, in more complex creatures this capacity develops during their life. Studying systems that would develop such an ability is a necessary and natural step towards the construction of robots showing more complex capacities.

Why should robots communicate?

Communication is a very desirable skill for robots to possess. We expect communicating robots to be able to carry out a wider range of tasks than is possible for dumb robots. For example, tasks requiring coordination within a group of robots would be easier if the robots were able to inform one another about what they were doing. It will also provide us with a way to teach robots and having them developing new skills after they have already been built. Communication abilities are bidirectional. Robots' ability to communicate could also provide us with a way to understand their internal function, by them explaining it.

Why learning to communicate?

Whilst fixed communication protocols can be designed, we expect that a flexible language which is acquired through a learning process will provide more adaptability towards the tasks the robots can carry out, the types of robots engaged in those tasks, and the user teaching the robots.

We study how communication abilities can be learned as part of general learning abilities.

  • Grounding Communication in social, embeded agents. Aude Billard and Kerstin Dautenhahn. To appear in the proceedings of TIMR UK conference , Buxton 4.9.97 - 5.9.1997.
  • 'Robot's first steps, Robot's first words ...'. Aude Billard and Gillian Hayes. In the proceedings of GALA'97 conference, Edinburgh, 4-6 April 97.
  • Home pages of people involved in this research:

  • Aude Billard
  • Dr. Gillian Hayes
  • Dr. Kerstin Dautenhahn