In this paper, we describe an approach to adapting home service robot behaviors by reusing the experience learned and interacting with humans. Experience is situation-based and is integrated into each behavior in a distributed way. Learning by examples is not only used to show the robot tasks but also to interact with humans. Object constraints in the demonstrated example are extracted by virtual object movement in image space. The experience gained for each individual behavior affects the whole action sequence of conducting a task. Inspired by the way of the prelinguistic infants learn and interact with their parents, demonstrations are used to interact with robots to adapt their behaviors. Experiments on a real physical robot system have successfully tested the approach.
展开▼
