Laser Scan Matching in Polar Coordinates Using Gaussian Process
Laser scan matching serves an important role as pose estimation in autonomous navigation for mobile robots.As a classic approach,iterative closest point(ICP)finds corresponding point pairs in a brute-force way,which is time consuming.For eliminating the cost of searching the correspondence between the point pairs,polar scan matching(PSM)uses a matching bearing rule by making use of the laser measurements collected in polar coordinates.However,PSM might occur mismatching when the irrelevant region of the reference and current scans has a similar distribution on polar range.In order to obtain better results,we propose a novel laser scan matching method based on a new type of map representation using GP regression in polar coordinates.With this map representation,the corresponding point pairs can be found in a simple and efficient way.Based on these pairs,we get the robot pose by iteratively estimating the orientation and translation.For dense-point-scan data sets,our approach demonstrates an outstanding performance compared with traditional approaches like ICP and PSM in terms of both accuracy and efficiency.
Laser scan matching Gaussian process Polar coordinates
Yinqiang Wang Bo Li Bo Han Yu Zhang Wenjie Zhao
Zhejiang University,Hangzhou,China
国际会议
江苏镇江
英文
106-115
2019-09-20（万方平台首次上网日期，不代表论文的发表时间）