A novel Loop Closure Approach in 2D LIDAR SLAM

Simultaneous Localization and Mapping (SLAM) technology has made great progress. However, in the unstructured environment, loopback detection is still a challenging task, especially in 2D laser SLAM. The scanning frame of 2D lidar carries less information, and it is difficult to extract features. Most of the current solutions are direct point cloud matching, which has the disadvantage of heavy computation. To solve this problem, a loop-back detection method based on Scan Context and principal component analysis (PCA) is proposed in this paper. The method constructs the scanning Context of laser scanning frame, reduces its dimension, and uses the feature vectors after dimensionality reduction to build KD tree to accelerate loop-back detection. This method can reduce the consumption of computing resources while ensuring the accuracy of loopback detection. In addition, we implement a new 2D laser SLAM framework based on the loopback detection method. Experimental results show that the loopback detection method has good robustness in both indoor and outdoor environments, and has very low requirements on hardware resources. It can also achieve high loopback detection frequency in embedded systems.