抄録
This paper describes a localization method for a mobile robot that automatically corrects the accumulated positioning errors by using ground images recorded in the first run. The error correction is carried out in a laptop PC with low-memory consumption by using a quadtree space partitioning in the robot which is equipped with only a single camera for the positioning. The experimental results show that the trajectory estimated by tracking ground images was successfully corrected by the absolute search using the recorded images. The proposed method is available for achieving a mobile robot based on teaching and playback.
