複数ビークルの相対姿勢推定のためのLiDAR協調スキャンマッチング

抄録

This paper presents a relative pose estimation method for multiple vehicles that involves scan matching using light detection and ranging (LiDAR) measurements captured by neighboring vehicles. Such cooperative scan matching can be used to cooperative moving-object tracking by vehicles in global navigation satellite systems (GNSS)-denied environments. Each vehicle is equipped with a LiDAR and detects static measurements originating from static objects, such as building walls and utility poles, from its own LiDAR scan images by applying an occupancy grid method. Each vehicle then obtains point feature histograms related to static features using fast point feature histograms (FPFH). The information related to LiDAR measurements and point feature histograms is sent to a central server. After the central server matches the environmental features obtained by the vehicles based on point feature histograms and estimates the relative pose of the vehicles by random sample consensus (RANSAC) based coarse registration and normal distributions transform (NDT) scan matching-based fine registration. The performance of the proposed method is evaluated using experimental results obtained in two road environments.