Lidar-Artificial-Marker Odometry for a Surface Climbing Robot via Factor Graph

This paper presents a tightly-coupled Lidar-Artificial-Marker odometry algorithm for a surface climbing robot via factor graph, offering robust self-localization performance. Feature degradation leads to lower accuracy of Lidar odometry pose estimation in structure-less environments. Therefore, we use artificial marker such as tag to improve localization accuracy. However, the estimated tag pose decreases in accuracy when the angle and distance become larger, so we use lidar point cloud to correct the tag pose which used as initial estimate of lidar frame pose. More corners are extracted from tag combining with lidar feature points to calculate the highly precise pose of unmanned vehicles. And one strategy is proposed to use estimated tag pose to directly estimate unmanned vehicles pose according the precision of tag pose, which can improve the estimated speed. The effectiveness and real-time performance of the proposed algorithm were investigated by experiments of unmanned vehicle.