ICP based on Polar Point Matching with application to Graph-SLAM

To solve the range scan matching problem, this paper describes a matching rule for points in polar coordinate, i.e. PPMR (Polar Point Matching Rule), which takes into account both rotation and translation error of the scanner displacement at the same time, and can be applied for both 2D and 3D scans. Based on PPMR, a fast correspondence search algorithm is developed taking advantage of the structure of laser scan data. The final scan matching algorithm, PmICP ICP (Iterative Closest/Corresponding Point) based on Polar Point Matching, is tested on several 2D range datasets. Experiments suggest the algorithm is accurate and fast, which guarantees the real-time application. In the end, this paper proposes a method which selects key scans from scan sequence and finds the best-match key scan for every new scan according to some evaluation functions. This method works online and can be incorporated into Graph-SLAM (Simultaneous Localization and Mapping).