Sensory fusion and emerging behaviours in an anthropomorphic robot as a man-machine interface

Scientists involved in developing human-machine interfaces have recently begun to address complex problems such as biological believability and how to design machines which can perceive, learn and make choices. Once this ambitious target has been reached, dynamic human-machine interactions will have to be studied focusing in particular on possible changes in the interactive structure and thus on any manifestations of emerging behaviours. The FACE (Facial Automaton for Conveying Emotions) project addresses both issues. FACE is a life-like artifact intended as a believable human-machine interface that is able to engage in nonverbal communication by imitating and learning the emotional behaviour of an interlocutor. This paper outlines the biomimetic systems it is equipped with, the main focus being on neural control architecture which consists of a sensory-motor map that can permit sensory fusion plus a neurocontroller able to navigate within a simplified behavioural space. The neurocontroller is based on recent discoveries of the role of astrocytes in cognitive processes.