抄録
This paper describes a human-robot cooperation system which enables robust behaviors of robots in a real world by combining human's ability of recognition, inference and decision with robot's autonomy. In previous systems, interactions between operators and robots are designed by programmers. We consider that the interactions should dynamically change through iterative process of human-robot cooperation. In the system, the robots obtain autonomous behaviors by memorizing human's instructions and obtain autonomous behaviors for behaving in complex environments. We have designed a system which consists of vision guided mobile robot, several cameras embedded in the environment and visual information interface with an idea of layered interaction structure, and we have experimented in a town model. We call this system DVScoop (Distributed Vision System for human-robot cooperation) .
