In urban dynamic environment, most of the existing works on LiDAR SLAM are based on static scene assumption and are greatly affected by dynamic obstacles. In order to solve this problem, this paper is based on F-LOAM, and adopts FA-RANSAC algorithm, improved ScanContext algorithm and global optimization to propose a robust and fast LiDAR Odometry and Mapping (RF-LOAM). Firstly, the Region Growing algorithm is used to cluster the fan-shaped grids. Then, we propose FA-RANSAC algorithm base on feature information and adaptive threshold for dynamic objects removal, and extracts the static edge and planar feature points for the first distortion compensation. Afterward, estimated pose is calculated by the static feature points and is used to perform the second distortion compensation. Then, the height difference and adaptive distance threshold are used to improve the accuracy of ScanContext, and the efficiency of ScanContext is improved by deleting the loop closure historical matching frames and simplifying the feature matching. Finally, global optimization is used for keyframe. The experimental tests are carried out on the KITTI datasets, Urbanloco datasets and our Extracted dataset. The results show that compared with the state-of-the-art SLAM methods, our method can not only accurately complete dynamic objects removal and loop closure detection, but also achieve more robust and faster localization and mapping in urban dynamic scenes.
24 August 2021 in IET Intelligent Transport Systems
Adjacent obstacles are difficult to be distinguished, and remote obstacle is easily detected as multiple targets. Besides, occluded obstacles are difficult to track, and tracking velocity of the obstacles in rapid acceleration or deceleration converges slowly in the urban environment. In view of this, an obstacle detection and tracking method based on multi‐lidar is proposed. Firstly, based on the vehicle kinematics model, motion compensation is adopted to solve the space‐time synchronization problem among lidars after road segmentation, and data level fusion is completed. Next, the obstacles are detected by combining adaptive voxel grid DBSCAN (AVG‐DBSCAN) algorithm and Region Growing (RG) algorithm. Then, the Breadth First Search (BFS) algorithm and KD tree are adopted to improve the Evolutionary Hungarian algorithm for the fast association of obstacles. Finally, the occlusion rate and dynamic region selection are used to track obstacles accurately, based on the Kalman Filter of uniform acceleration model. The experimental results on the authors' extracted urban dataset show that the proposed method can effectively fuse multi‐lidar data and outperforms other methods in obstacle detection and tracking. Its average accuracy for detection is 97.53% and its average accuracy for tracking is 95.1%. The average duration of the entire process is only 30 ms.
28 April 2025 in IEEE Transactions on Instrumentation and Measurement
In order to improve the accuracy and efficiency of simultaneous localization and mapping (SLAM) in large, degraded, and complex areas, real-time multi-robot cooperative SLAM has more obvious advantages in accuracy, fault tolerance, and flexibility than single robot SLAM. However, currently existing cooperative SLAM algorithms have the following issues. On the one hand, poor localization of degraded scenes and incomplete consideration of inter-robot loop constraints lead to insufficient robustness. On the other hand, the bandwidth occupation is large in inter-robot loop areas. Therefore, in order to solve the above problems, we innovatively propose the distributed robust and communication-efficient multi-robot cooperative LiDAR-inertial SLAM (DRCM-CSLAM). Firstly, the FAST-LIO2 with loop detection and loop correction is integrated into cooperative framework to improve the robustness in degraded scenes. Subsequently, a two-stage loop filtering is proposed to improve the accuracy of the two-stage optimization of DiSCo-SLAM by fully utilizing inter-robot loop constraints. Finally, we are the first to combine Scan Context (SC) descriptor and incremental octree to design a lightweight and efficient communication strategy, significantly reducing bandwidth occupation of inter-robot loop areas and ensuring real-time performance. The experiments are tested on KITTI datasets and collected datasets. The results show that our method has the superior performance in robustness, bandwidth and runtime. Compared with DiSCo-SLAM, our method reduces absolute translational error (RMSE) by 42.9% and bandwidth by 90%.
Adjacent obstacles are difficult to be distinguished, and remote obstacles are detected easily to split. Besides, limited deep learning samples easily result in missed detection of obstacles in urban environment. In view of this, a fast and robust detection method is proposed by fusing Double-Layer Region Growth algorithm and Grid-SECOND detector. At first, SECOND detector is improved by replacing voxel grids with 2D grids and adopting multi-dimensional features to detect obstacles, which can reduce the time consumption and ensure the accurate detection of remote obstacles. Then, the first Region Growing algorithm is used to cluster the undetected and non-empty grids, which can detect obstacles outside the training set. At last, the second Region Growing algorithm is used to refine the detection results of obstacles with larger volume and multimulti multimulti-obstacles obstacles obstacles grids, and complete the obstacle detection. Through testing in our extracted urban dataset and KITTI dataset, it is verified that the proposed method outperforms state-of-the-art methods and can accurately achieve obstacle detection. The average duration of the entire process is about 50ms.
20 August 2025 in IEEE Transactions on Instrumentation and Measurement
Underground sheltered spaces, such as utility tunnels, mines, railways, are characterized by unique features including enclosure, complexity, and structural diversity, which present great challenges to collaborative perception and localization in clustered unmanned systems. This paper firstly analyzes the current research of clustered unmanned systems in underground spaces. Secondly, Considering the challenges of communication limitation, degraded environments, constrained perception in underground spaces, this paper elaborates on the progress of key technologies for collaborative localization and perception in unmanned systems, including data exchange methods for robotic cluster under communication constraints, robust and resilient SLAM algorithms under degraded environments, collaboration strategies for heterogeneous multi-robot systems in underground environments and collaborative semantic SLAM under limited environmental perception. Finally, the technological bottlenecks and development trends of clustered unmanned systems in underground environments are summarized and discussed.
Multirobot collaborative SLAM, as a key direction in Internet of Things (IoT) applications, faces the critical challenge of achieving accurate self-pose and mutual-pose estimation. Existing distributed pose optimization methods often suffer from unbalanced computational allocation and insufficient utilization of inter-robot information. To address these limitations, we propose a balanced and robust distributed pose optimization (BR-DPO) framework that integrates three complementary layers: inter-robot global transformation, intra-robot factor graph, and real-time pose estimation. First, a flexible load-aware bidirectional global pose optimization (FLB-GPO) module is designed to achieve inter-robot adaptive computational balance. Next, a multi-source evaluated local pose optimization (ME-LPO) module is introduced, which allows high-precision robots to assist low-precision robots for accurate intra-robot factor graph refinement on different scenes. Then, a geometric observation-driven degenerate pose correction (GO-DPC) module is developed to improve observation precision and real-time pose estimation accuracy in degraded environments. Together, these modules establish a hierarchical optimization architecture that achieves inter-robot computational balance, intra-robot refined pose, and real-time degeneration correction. Both simulation and real-world experiments demonstrate that the proposed framework consistently outperforms state-of-the-art methods in terms of computational balance and robustness.