Vision Based Working Robot Based on Architecture of Control with Error Recovery.

We have developed a robot that inspects the air volume from square air diffusers that are called "anemo" on the ceiling at office buildings. The robot moves automatically, relocating its self-position by recognizing the anemo as the landmark with CCD cameras. However, self-position of the robot cannot be calculated, when the robot fails in recognizing the landmark. Then the robot stops working. Therefore, we have proposed the behaviors of error recovery. In this paper, we propose "Hierarchical Adaptive and Learning Architecture System(HALAS)" that makes the robot to be adapted to the environment by learning so that the robot can continue to work.