Self-supervised learning of LiDAR odometry based on spherical projection

Fu, Xu; Liu, Cong; Zhang, Chengjin; Sun, Zihao; Song, Yun Heub; Xu, Qingyang; Yuan, Xianfeng

doi:10.1177/17298806221078669

Cited by 7 publications

(2 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…PWCLO-Net [25] and SVDLO [46] attempt to process the raw 3D point clouds by PointNet-based models [47]. For the training loss under selfsupervision, ICP-like [48][49][50], point-to-point matching loss [51], point-to-plane matching loss [18,20], and plane-to-plane matching loss [20] are extensively used.…”

Section: Lidar Odometrymentioning

confidence: 99%

UnVELO: Unsupervised Vision-Enhanced LiDAR Odometry with Online Correction

et al. 2023

Sensors

View full text Add to dashboard Cite

Due to the complementary characteristics of visual and LiDAR information, these two modalities have been fused to facilitate many vision tasks. However, current studies of learning-based odometries mainly focus on either the visual or LiDAR modality, leaving visual–LiDAR odometries (VLOs) under-explored. This work proposes a new method to implement an unsupervised VLO, which adopts a LiDAR-dominant scheme to fuse the two modalities. We, therefore, refer to it as unsupervised vision-enhanced LiDAR odometry (UnVELO). It converts 3D LiDAR points into a dense vertex map via spherical projection and generates a vertex color map by colorizing each vertex with visual information. Further, a point-to-plane distance-based geometric loss and a photometric-error-based visual loss are, respectively, placed on locally planar regions and cluttered regions. Last, but not least, we designed an online pose-correction module to refine the pose predicted by the trained UnVELO during test time. In contrast to the vision-dominant fusion scheme adopted in most previous VLOs, our LiDAR-dominant method adopts the dense representations for both modalities, which facilitates the visual–LiDAR fusion. Besides, our method uses the accurate LiDAR measurements instead of the predicted noisy dense depth maps, which significantly improves the robustness to illumination variations, as well as the efficiency of the online pose correction. The experiments on the KITTI and DSEC datasets showed that our method outperformed previous two-frame-based learning methods. It was also competitive with hybrid methods that integrate a global optimization on multiple or all frames.

show abstract

Section: Lidar Odometrymentioning

confidence: 99%

UnVELO: Unsupervised Vision-Enhanced LiDAR Odometry with Online Correction

et al. 2023

Sensors

View full text Add to dashboard Cite

show abstract

“…The odometry system requires a rotary encoder sensor to detect the number of wheel rotations [28] [29]. Area mapping using the odometry method is to estimate changes in the robot's position over time in a Cartesian diagram [30] [31]. The result is data on the coordinates and direction of the robot [32][33] [34].…”

Section: Introductionmentioning

confidence: 99%

Application of Odometry and Dijkstra Algorithm as Navigation and Shortest Path Determination System of Warehouse Mobile Robot

Ubaidillah,

Sukri

2023

Journal of Robotics and Control

View full text Add to dashboard Cite

One of the technologies in the industrial world that utilizes robots is the delivery of goods in warehouses, especially in the goods distribution process. This is very useful, especially in terms of resource efficiency and reducing human error. The existing system in this process usually uses the line follower concept on the robot's path with a camera sensor to determine the destination location. If the line and destination are not detected by the sensor or camera, the robot's navigation system will experience an error. it can happen if the sensor is dirty or the track is faded. The aim of this research is to develop a robot navigation system for efficient goods delivery in warehouses by integrating odometry and Dijkstra's algorithm for path planning. Holonomic robot is a robot that moves freely without changing direction to produce motion with high mobility. Dijkstra's algorithm is added to the holonomic robot to obtain the fastest trajectory. by calculating the distance of the node that has not been passed from the initial position, if in the calculation the algorithm finds a shorter distance it will be stored as a new route replacing the previously recorded route. the distance traversed by the djikstra algorithm is 780 mm while a distance of 1100 mm obtains the other routes. The time for using the Djikstra method is proven to be 5.3 seconds faster than the track without the Djikstra method with the same speed. Uneven track terrain can result in a shift in the robot's position so that it can affect the travel data. The conclusion is that odometry and Dijkstra's algorithm as a planning system and finding the shortest path are very efficient for warehouse robots to deliver goods than ordinary line followers without Dijkstra, both in terms of distance and travel time.

show abstract

A novel indoor localization algorithm based on a modified EKF using virtual dynamic point landmarks for 2D grid maps

Altınpınar,

Sezer

2023

Robotics and Autonomous Systems

View full text Add to dashboard Cite

Self-supervised learning of LiDAR odometry based on spherical projection

Cited by 7 publications

References 23 publications

UnVELO: Unsupervised Vision-Enhanced LiDAR Odometry with Online Correction

UnVELO: Unsupervised Vision-Enhanced LiDAR Odometry with Online Correction

Application of Odometry and Dijkstra Algorithm as Navigation and Shortest Path Determination System of Warehouse Mobile Robot

A novel indoor localization algorithm based on a modified EKF using virtual dynamic point landmarks for 2D grid maps

Contact Info

Product

Resources

About