Efficient Dynamic LiDAR Odometry for Mobile Robots with Structured Point Clouds

TL;DR

This work proposes a real-time dynamic LiDAR odometry pipeline for mobile robots in Urban Search and Rescue (USAR) scenarios and reuse data between the odometry and detection module to enhance efficiency on computationally limited robots.

Abstract

We propose a real-time dynamic LiDAR odometry pipeline for mobile robots in Urban Search and Rescue (USAR) scenarios. Existing approaches to dynamic object detection often rely on pretrained learned networks or computationally expensive volumetric maps. To enhance efficiency on computationally limited robots, we reuse data between the odometry and detection module. Utilizing a range image segmentation technique and a novel residual-based heuristic, our method distinguishes dynamic from static objects before integrating them into the point cloud map. The approach demonstrates robust object tracking and improved map accuracy in environments with numerous dynamic objects. Even highly non-rigid objects, such as running humans, are accurately detected at point level without prior downsampling of the point cloud and hence, without loss of information. Evaluation on simulated and real-world data validates its computational efficiency. Compared to a state-of-the-art volumetric method, our approach shows comparable detection performance at a fraction of the processing time, adding only 14 ms to the odometry module for dynamic object detection and tracking. The implementation and a new real-world dataset are available as open-source for further research.

Figures (9)

• Figure 2: Application of the proposed approach for dynamic LiDAR odometry, enabling the efficient segmentation and tracking of highly articulated objects. Top: A group of jumping persons is detected and tracked in a highly dynamic sequence. Dynamic points are shown in green, and trajectories are indicated in blue. Bottom: The range image used for object segmentation of the above cloud. Different colors represent different static and dynamic objects.
• Figure 3: System overview. Two consecutive scans are registered using the odometry module (blue). The detection module (green) segments the projected range image and residual image into individual objects. The Tracking and Association module (yellow) assigns a dynamic state to each object. Non-static points are removed from the current scan before integrating it into the keyframe database. Ghost traces are removed from the global map (red) based on the objects' bounding boxes.
• Figure 4: Residual image as obtained from the GICP algorithm. For better visibility, the registration was performed on the original point cloud, whereas the method uses the downsampled and filtered cloud. A rainbow color map highlights the points with higher residuals. Black pixels represent invalid points.
• Figure 5: Dynamic state update. At each time step, all tracked objects are updated and can change their dynamic state based on the number of detections (hits), their average residuum and the displacement from their origin.
• Figure 6: IoU, precision, and recall on the DOALS small town simulation sequence for different values of the residuum threshold (see Eq. \ref{['eqn:residual_condition']}).