Abstract: | Roboticists have already invested considerable energy in modelling on robots the learning capacities of single animals. However, the field of learning by imitation, in which one agent learns from perceiving the behaviour of another, is largely untouched. The range of behaviour described as "learning by imitation" in the ethological literature is wide. We argue that building and observing robots which instantiate the mechanisms hypothesised to underlie these types of behaviour will cut through the chaos of the terminology and illuminate those mechanisms with explanatory adequacy. We present an architecture within which to study one of these mechanisms and describe two implementations, a simulator and a robot, with which we are studying different aspects of the mechanism.
|