Embedding Robotic Agents in the Social Environment

This paper discusses the interactive vision approach, which advocates using knowledge from the human sciences on the structure and dynamics of human-human interaction in the development of machine vision systems and interactive robots. While this approach is discussed generally, the particular case of the system being developed for the Aurora project (which aims to produce a robot to be used as a tool in the therapy of children with autism) is especially considered, with description of the design of the machine vision system being employed and discussion of ideas from the human sciences with particular reference to the Aurora system. An example architecture for a simple interactive agent, which will likely form the basis for the first implementation of this system, is briefly described and a description of hardware used for the Aurora system is given.