Wenlu Yu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Wenlu Yu has a total of 12 co-authors. Here are the top co-authors by number of shared publications:
- Jie Xu
Nanyang Technological University
3 shared publications - Lijun Zhao
Harbin Institute of Technology
3 shared publications - Ruifeng Li
Harbin Institute of Technology
3 shared publications - Song Huang
State Key Laboratory of Robotics and Systems, Harbin Institute of Technology, Harbin, China
2 shared publications - Bo Zhang
State Key Laboratory of Robotics and Systems, Harbin Institute of Technology, Harbin, China
1 shared publication - Lihua Xie
Nanyang Technological University
1 shared publication
Latest Publications
- 29 July 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 11 (23) , 37562-37570
M-DIVO: Multiple ToF RGB-D Cameras-Enhanced Depth–Inertial–Visual Odometry
Time of flight (ToF) RGB-D cameras provide a wealth of information for SLAM systems. However, the limited field of view (FOV) of a single ToF RGB-D camera and the small range of its depth measurement module make it prone to degeneracy when relying solely on visual or depth information for SLAM, a problem typical of unimodal SLAM algorithms. To address this issue, this paper presents M-DIVO: an IEKF-based odometry that fuses visual, depth (similar to LiDAR), and inertial modules from multiple ToF RGB-D cameras. It comprises two direct method subsystems: the depth inertial odometry (DIO) subsystem, which constructs point-to-plane constraints from multiple depth modules, and the visual-inertial odometry (VIO) subsystem, which optimizes pose using photometric error constructed by multiple cameras. Additionally, to manage the significant computational load from processing multiple sensors and multimodal information, we introduce a multimodal redundancy scheduling mechanism (MRSM): prioritizing the DIO subsystem with the VIO subsystem as auxiliary, executing the VIO subsystem only when degeneracy occurs in the DIO subsystem. We also propose a “External First, Internal Last" strategy for calibrating multiple external and internal sensors. Experiments demonstrate that compared to unimodal SLAM, our method achieves higher robustness and precision, as well as satisfactory real-time performance. The proposed calibration strategy is demonstrated to be more accurate than the traditional IMU-centric approach. The code is open source.
- 29 July 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 11 (23) , 37562-37570
M-DIVO: Multiple ToF RGB-D Cameras-Enhanced Depth–Inertial–Visual Odometry
Time of flight (ToF) RGB-D cameras provide a wealth of information for SLAM systems. However, the limited field of view (FOV) of a single ToF RGB-D camera and the small range of its depth measurement module make it prone to degeneracy when relying solely on visual or depth information for SLAM, a problem typical of unimodal SLAM algorithms. To address this issue, this paper presents M-DIVO: an IEKF-based odometry that fuses visual, depth (similar to LiDAR), and inertial modules from multiple ToF RGB-D cameras. It comprises two direct method subsystems: the depth inertial odometry (DIO) subsystem, which constructs point-to-plane constraints from multiple depth modules, and the visual-inertial odometry (VIO) subsystem, which optimizes pose using photometric error constructed by multiple cameras. Additionally, to manage the significant computational load from processing multiple sensors and multimodal information, we introduce a multimodal redundancy scheduling mechanism (MRSM): prioritizing the DIO subsystem with the VIO subsystem as auxiliary, executing the VIO subsystem only when degeneracy occurs in the DIO subsystem. We also propose a “External First, Internal Last" strategy for calibrating multiple external and internal sensors. Experiments demonstrate that compared to unimodal SLAM, our method achieves higher robustness and precision, as well as satisfactory real-time performance. The proposed calibration strategy is demonstrated to be more accurate than the traditional IMU-centric approach. The code is open source.
- 25 January 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 9 (3) , 2590-2597
LiDAR-Link: Observability-Aware Probabilistic Plane-Based Extrinsic Calibration for Non-Overlapping Solid-State LiDARs
As solid-state LiDAR technology advances, mobile robotics and autonomous driving increasingly rely on multiple solid-state LiDARs for perception. However, limited or non-overlapping fields of view (FoV) among these sensors pose significant challenges for extrinsic calibration. Moreover, there are no quantitative indicators to evaluate calibration quality currently. To tackle these challenges, we introduce LiDAR-Link, a novel approach for calibration and evaluation, consisting of LiDAR-Bridge and LiDAR-Align. LiDAR-Bridge uses a wide-angle LiDAR to act as an intermediary, aligning point clouds from small-angle solid-state LiDARs. These small LiDARs have limited or non-overlapping FoVs, allowing us to indirectly determine the extrinsics between them. LiDAR-Align adaptively constructs a Voxel Probabilistic Plane Map to efficiently match point clouds and differentiate the contributions of various high-quality planar features. To align these point clouds, Iterated Extended Kalman Filter (IEKF) is utilized, which is based on point-to-plane residuals. The resulting covariance verifies whether point cloud alignment constraints are met and assesses calibration reliability. Furthermore, LiDAR-Align supports multi-scene joint calibration to overcome the limitations of fewer constraints caused by smaller LiDAR FoVs. We validate the accuracy of our method and the significance of observability analysis through a comprehensive set of experiments.
- 25 January 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 9 (3) , 2590-2597
LiDAR-Link: Observability-Aware Probabilistic Plane-Based Extrinsic Calibration for Non-Overlapping Solid-State LiDARs
As solid-state LiDAR technology advances, mobile robotics and autonomous driving increasingly rely on multiple solid-state LiDARs for perception. However, limited or non-overlapping fields of view (FoV) among these sensors pose significant challenges for extrinsic calibration. Moreover, there are no quantitative indicators to evaluate calibration quality currently. To tackle these challenges, we introduce LiDAR-Link, a novel approach for calibration and evaluation, consisting of LiDAR-Bridge and LiDAR-Align. LiDAR-Bridge uses a wide-angle LiDAR to act as an intermediary, aligning point clouds from small-angle solid-state LiDARs. These small LiDARs have limited or non-overlapping FoVs, allowing us to indirectly determine the extrinsics between them. LiDAR-Align adaptively constructs a Voxel Probabilistic Plane Map to efficiently match point clouds and differentiate the contributions of various high-quality planar features. To align these point clouds, Iterated Extended Kalman Filter (IEKF) is utilized, which is based on point-to-plane residuals. The resulting covariance verifies whether point cloud alignment constraints are met and assesses calibration reliability. Furthermore, LiDAR-Align supports multi-scene joint calibration to overcome the limitations of fewer constraints caused by smaller LiDAR FoVs. We validate the accuracy of our method and the significance of observability analysis through a comprehensive set of experiments.
- 19 September 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 7 (4) , 11879-11885
CamMap: Extrinsic Calibration of Non-Overlapping Cameras Based on SLAM Map Alignment
Multiple cameras have emerged as a promising technology for robots and vehicles due to their broad fields of view (FoV) and high resolution. However, there are often limited or no overlapping FoVs among cameras, bringing challenges to estimating extrinsic camera parameters. To overcome this problem, we propose CamMap: a novel 6-degree-of-freedom (DoF) extrinsic calibration pipeline. Following three operating rules, we make a multi-camera rig capture some similar image sequences individually to create sparse feature-based maps with a SLAM system. A two-stage optimization problem is formulated to align the maps and obtain the transformations between them based on bidirectional reprojection. The transformations are exactly the extrinsic parameters. Supporting diverse camera types, the pipeline is available in any texture-rich environment. It can calibrate any number of cameras simultaneously without requiring calibration patterns, synchronization, same resolution and frequency. The pipeline is evaluated on cameras with limited and no overlapping FoVs. In the experiments, we demonstrate our method's accuracy and efficiency. The absolute pose error (APE) between Kalibr and CamMap is less than 0.025.
- 19 September 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 7 (4) , 11879-11885
CamMap: Extrinsic Calibration of Non-Overlapping Cameras Based on SLAM Map Alignment
Multiple cameras have emerged as a promising technology for robots and vehicles due to their broad fields of view (FoV) and high resolution. However, there are often limited or no overlapping FoVs among cameras, bringing challenges to estimating extrinsic camera parameters. To overcome this problem, we propose CamMap: a novel 6-degree-of-freedom (DoF) extrinsic calibration pipeline. Following three operating rules, we make a multi-camera rig capture some similar image sequences individually to create sparse feature-based maps with a SLAM system. A two-stage optimization problem is formulated to align the maps and obtain the transformations between them based on bidirectional reprojection. The transformations are exactly the extrinsic parameters. Supporting diverse camera types, the pipeline is available in any texture-rich environment. It can calibrate any number of cameras simultaneously without requiring calibration patterns, synchronization, same resolution and frequency. The pipeline is evaluated on cameras with limited and no overlapping FoVs. In the experiments, we demonstrate our method's accuracy and efficiency. The absolute pose error (APE) between Kalibr and CamMap is less than 0.025.