Human-robot communication by physical interaction - distributed agent control system and the learning algorithm

This study aims to clarify the cooperation intelligence of robots with the emotion model referring to the human biological system. This paper describes the functions of the arm system of the behavior-based robot, Wamoeba-2, which has an emotion model and can communicate with human. The specifications of the arm system were designed based on the consideration of the human robot physical interaction. Each joint of the arm is equipped with a torque sensor, and the arm is controlled by a distributed agent network system. The network architecture is acquired in a neural network by the feedback-error-learning algorithm. In the experiment, human beings can play with the arm system by physical interaction.