複数Lidarによる協調移動物体追跡のための環境特徴を用いた協調スキャンマッチング

抄録

This paper describes a cooperative tracking (estimation of pose and size) of moving objects using multiple lidars mounted on vehicles (mobile sensor nodes). Although cooperative tracking improves tracking accuracy and reliability by sharing sensing data among nearby sensor nodes, the sensor nodes have to accurately identify their relative poses to perform a reliable cooperative tracking. In GNSS-denied environments, the relative pose is estimated by scan matching using lidar scan data captured in the overlapped sensor areas of sensor nodes. Such cooperative scan matching is performed by two registrations: coarse registration using Fast Point Feature Histograms (FPFH) and Sample Consensus Initial Alignment (SACIA) and fine registration using Normal Distributions Transform (NDT) scan matching. The experimental results by two mobile sensor nodes validate the proposed method.