Affordance based imitation bootstrapping with motionese

Learning through self-exploration and imitation are crucial mechanisms in developing sensorimotor skills for human infants. In our previous work, we showed that a robot can self-discover behavior primitives and learn object affordances similar to infants. Then building predictive mechanisms at the affordance level allowed movement planning and simple goal-level imitation on our robot. The work we describe in this paper builds upon this system; it describes how our robot, by using the learned behaviors and prediction mechanisms, can go beyond simple goal-level imitation and become a better imitator. For this, we develop mechanisms to enable the robot to recognize and segment, with the help of the demonstrator, an ongoing action in terms of its affordance based goal satisfaction. Extracting sub-goals or important features from a demonstration is not straightforward as demonstrated action trajectory may not correspond to any robot behavior developed so far. Inspired from infant development, we use motionese to enable the robot to identify the important steps and the boundaries in the otherwise complex stream of motion involving multi-objects. Once the sub-goals are obtained, the robot imitates the observed action by chaining these sub-goals and satisfying them sequentially. In the experiments, a tutor used his own action repertoire to move an object around, and the robot was able to detect and achieve the sub-goals with its affordance prediction mechanisms and behavior primitives.