Kindergarten SLAM utilizing laser range sensor and retro-reflective markers

For realization of robust SLAM available for dynamic environment where active children walk around in the kindergarten, a simplified SLAM using virtual comb-shaped markers is suggested in this study. To easily extract the feature point for the simplified SLAM, the virtual comb-shape markers were devised by using reflective materials, and the positions of the marker were detected using a later range finder and Hough transformation. Combining the marker detection method and extended Kalman filter technique, the simplified SLAM was constructed. From the experiment on the real environment using an actual omni-directional robot, it was shown that position of landmark (LM) was able to be computed with high accuracy by comparing with standard methods. Through the simulation of the simplified SLAM, it was confirmed that the average error between true value of the robot position and the other positional value estimated by the presented SLAM was 24 mm, which corresponds to error rate as small as 0.14 % against total running distance.