Reducing the cumulative error in the process of simultaneous localization and mapping (SLAM) has always been a hot issue. In this paper, in order to improve the localization and mapping accuracy of ground vehicles, we proposed a novel optimized lidar odometry and mapping method using ground plane constraints and SegMatch-based loop detection. We only used the lidar point cloud to estimate the pose between consecutive frames, without any other sensors, such as Global Positioning System (GPS) and Inertial Measurement Unit (IMU). Firstly, the ground plane constraints were used to reduce matching errors. Then, based on more accurate lidar odometry obtained from lidar odometry and mapping (LOAM), SegMatch completed segmentation matching and loop detection to optimize the global pose. The neighborhood search was also used to accomplish the loop detection task in case of failure. Finally, the proposed method was evaluated and compared with the existing 3D lidar SLAM methods. Experiment results showed that the proposed method could realize low drift localization and dense 3D point cloud map construction.
Existing point-based, sparse voxel-based, or hybrid point cloud processing methods require time-consuming neighborhood searches or sparse 3D convolutions, which consume a lot of time and computational resources. Therefore, it is difficult to run at high speed in real time on mobile devices. To this end, we reconstructed the internal structure based on RangeNet++1 and designed an efficient and lightweight network named AMBrnet, still using the encoder-decoder architecture. An asymmetric multi-branch aggregation module is designed to directly aggregate the input information and provide rich information for subsequent step encoding. The dual-branch structure is introduced to combine with the encoder and decoder to strengthen the information encoding and decoding capabilities of the network. Weighted cross-entropy loss combined with Lovász-Softmax loss2 is used to directly optimize the Jaccard index (IoU). In the network inference stage, structural reparameterization is introduced to ensure that the inference speed is improved based on the same accuracy, reducing the number of network parameters. We evaluate the proposed model on the SemanticKITTI dataset. The prediction accuracy is better than most existing networks, notably its inference speed is up to 43.9 Hz. Experimental data show that AMBrnet is well suited for real-time high-speed point cloud processing on outdoor mobile devices.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.