Abstract
Recently, many studies have been conducted on robot positioning using self-localization method without GPS. One of the efficient methods is called SLAM, Simultaneous Localization And Mapping. This method can estimate robot position with high accuracy by using both internal sensors such as gyro sensor, rotary encoder, and external sensors such as camera sensor to get landmark information. In the SLAM method, robot have to find and detect some landmarks with own camera sensor, but it needs high performance CPU and wide angle camera lens system, and it became an issue for downsizing of robot. To tackle this issue, we propose a novel method called generalized measuring worm algorithm. Fundamental idea of this algorithm is mutual positioning using two robots. We applied a pair of robots for estimating position each other, and made maps using each robot's position as landmarks. Robots equip a two dimension maker that are known size and shape beforehand for distance and direction measuring. Our method assumes a situation of two robots set on line. We call a robot of backward mother robot, while forward child robot. At a first step of measuring, mother robot try to take a photo of child robot's 2D marker, calculates distance, and get direction information by using size and shape of marker. Next step, mother robot sends a command to child robot for correcting child robot direction by using information in mother robot's sight. After child robot correcting, mother robot moves toward child robot nearly. By repeating this operation, two robots can move to any destination. They also can estimate positioning; making maps of any coordinate system based mother robot's coordinate system. To validate effectively of our method, we made an experimental system, and we could show that our method can estimate position and make maps with less than 5% error.
