Loopback detection is the most critical part of the SLAM problem, which is directly related to the accuracy of the final output pose. During the long-term movement of the robot, the cumulative error continues to increase. When the robot returns to the previous position, new constraints can be added. The correct closed-loop information can be used to eliminate odometer errors, thereby obtaining a globally consistent map. Wrong closed loop will not only interfere with graph optimization, but also cause completely wrong graph construction results.
正在翻译中..