Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Degradation Resilient LiDAR-Radar-Inertial Odometry

Morten Nissov, Nikhil Khedekar, Kostas Alexis

Abstract

Enabling autonomous robots to operate robustly in challenging environments is necessary in a future with increased autonomy. For many autonomous systems, estimation and odometry remains a single point of failure, from which it can often be difficult, if not impossible, to recover. As such robust odometry solutions are of key importance. In this work a method for tightly-coupled LiDAR-Radar-Inertial fusion for odometry is proposed, enabling the mitigation of the effects of LiDAR degeneracy by leveraging a complementary perception modality while preserving the accuracy of LiDAR in well-conditioned environments. The proposed approach combines modalities in a factor graph-based windowed smoother with sensor information-specific factor formulations which enable, in the case of degeneracy, partial information to be conveyed to the graph along the non-degenerate axes. The proposed method is evaluated in real-world tests on a flying robot experiencing degraded conditions including geometric self-similarity as well as obscurant occlusion. For the benefit of the community we release the datasets presented: https://github.com/ntnu-arl/lidar_degeneracy_datasets.

Degradation Resilient LiDAR-Radar-Inertial Odometry

Abstract

Enabling autonomous robots to operate robustly in challenging environments is necessary in a future with increased autonomy. For many autonomous systems, estimation and odometry remains a single point of failure, from which it can often be difficult, if not impossible, to recover. As such robust odometry solutions are of key importance. In this work a method for tightly-coupled LiDAR-Radar-Inertial fusion for odometry is proposed, enabling the mitigation of the effects of LiDAR degeneracy by leveraging a complementary perception modality while preserving the accuracy of LiDAR in well-conditioned environments. The proposed approach combines modalities in a factor graph-based windowed smoother with sensor information-specific factor formulations which enable, in the case of degeneracy, partial information to be conveyed to the graph along the non-degenerate axes. The proposed method is evaluated in real-world tests on a flying robot experiencing degraded conditions including geometric self-similarity as well as obscurant occlusion. For the benefit of the community we release the datasets presented: https://github.com/ntnu-arl/lidar_degeneracy_datasets.
Paper Structure (23 sections, 14 equations, 5 figures, 3 tables)

This paper contains 23 sections, 14 equations, 5 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. LiDAR-based SLAM
  4. Radar-based SLAM
  5. Radar and LiDAR in Multi-modal SLAM
  6. Proposed Approach
  7. Notation
  8. State Estimation
  9. IMU Measurements
  10. LiDAR Measurements
  11. LiDAR Point-to-Plane Factor
  12. LiDAR Point-to-Line Factor
  13. Radar Measurement
  14. Radar Velocity Factor
  15. Uncertainty
  16. ...and 8 more sections

Figures (5)

  • Figure 1: Demonstration of the challenging environments explored in this paper, including a geometrically self-similar tunnel and a building with a fog-filled hallway. The effect of perceptual degradation is shown in the LiDAR point clouds. Radar data is fused to enable resilient odometry.
  • Figure 2: Architecture of the factor graph for the proposed method, including factors created from imu, LiDAR, and radar measurements.
  • Figure 3: Motion capture results comparing FAST-LIO2, the proposed method, and its versions without LiDAR (RIO) or radar (LIO).
  • Figure 4: Trajectories of the proposed method, the proposed method without LiDAR (RIO), the proposed method without radar (LIO), and FAST-LIO2 in the geometrically self-similar section of the Fyllingsdalen bicycle tunnel. Note FAST-LIO2 diverges backward after entering the self-similar region.
  • Figure 5: Left: Trajectories of the proposed method, the proposed method without radar (LIO), the proposed method without LiDAR (RIO), and FAST-LIO2 in the NTNU fog environment. Note FAST-LIO2 almost immediately diverges due to fog. Right: Third person camera view and views of the raw pointcloud of the LiDAR at Point A in the fog-filled hallway