Fine-registered Object LiDAR-inertial Odometry for a Solid-state LiDAR system

We propose the LiDAR-inertial odometry with object measurements for the solid-state LiDAR system with narrow field-of-view.Although the geometric feature has been used for the precise localization with LiDAR, the measurement vanishing can lead to the localization failure in the limited field-of-view.To address this problem, objects that are sufficiently present in a man-made environment can be used as localization measurements, but accurate registration is needed to formulate them due to partial observations.The point clouds in the object are registered and the processed measurements are coupled with the geometric measurements in the estimator.The effectiveness of the object measurements is verified through a virtual environment simulator, and the proposed algorithm shows superior localization performance compared to the case of geometric measurement alone.