Multimodal Sensors Fusion SLAM Based on Local Control and Multiscale Distance Analysis

Multimodal sensor fusion exhibits formidable advantages in simultaneous localization and mapping (SLAM). Nevertheless, integrating measurements from multimodal sensors at varying timestamps and frequencies can lead to the loss of certain state information, ultimately compromising performance. This article introduces a local control-based multisensors fusion algorithm that adeptly captures sensor data at different sampling rates and achieves spatial calibration, markedly enhancing the accuracy and real-time performance of extrinsic calibration. By leveraging a multiscale distance-based segmentation approach for loop closure detection, the algorithm continuously monitors and corrects accumulated trajectory errors, while fusing multimodal sensor data to bolster the accuracy and robustness of pose estimation in SLAM tasks. Experimental evaluations on the public NTU-VIRAL dataset reveal an overall precision improvement of 94.2%, 8.1%, and 86.6% over R3LIVE, CLIC, and FAST-LIVO, respectively. These results underscore the reliable accuracy and robustness of the proposed SLAM algorithm.