Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LIMOT: A Tightly-Coupled System for LiDAR-Inertial Odometry and Multi-Object Tracking

Zhongyang Zhu, Junqiao Zhao, Kai Huang, Xuebo Tian, Jiaye Lin, Chen Ye

TL;DR

LIMOT is proposed, a tightly-coupled multi-object tracking and LiDAR-inertial odometry system that is capable of accurately estimating the poses of both ego-vehicle and objects and achieves better pose and tracking accuracy than the previous work DL-SLOT and other baseline methods.

Abstract

Simultaneous localization and mapping (SLAM) is critical to the implementation of autonomous driving. Most LiDAR-inertial SLAM algorithms assume a static environment, leading to unreliable localization in dynamic environments. Moreover, the accurate tracking of moving objects is of great significance for the control and planning of autonomous vehicles. This study proposes LIMOT, a tightly-coupled multi-object tracking and LiDAR-inertial odometry system that is capable of accurately estimating the poses of both ego-vehicle and objects. We propose a trajectory-based dynamic feature filtering method, which filters out features belonging to moving objects by leveraging tracking results before scan-matching. Factor graph-based optimization is then conducted to optimize the bias of the IMU and the poses of both the ego-vehicle and surrounding objects in a sliding window. Experiments conducted on the KITTI tracking dataset and self-collected dataset show that our method achieves better pose and tracking accuracy than our previous work DL-SLOT and other baseline methods. Our open-source implementation is available at https://github.com/tiev-tongji/LIMOT.

LIMOT: A Tightly-Coupled System for LiDAR-Inertial Odometry and Multi-Object Tracking

TL;DR

LIMOT is proposed, a tightly-coupled multi-object tracking and LiDAR-inertial odometry system that is capable of accurately estimating the poses of both ego-vehicle and objects and achieves better pose and tracking accuracy than the previous work DL-SLOT and other baseline methods.

Abstract

Simultaneous localization and mapping (SLAM) is critical to the implementation of autonomous driving. Most LiDAR-inertial SLAM algorithms assume a static environment, leading to unreliable localization in dynamic environments. Moreover, the accurate tracking of moving objects is of great significance for the control and planning of autonomous vehicles. This study proposes LIMOT, a tightly-coupled multi-object tracking and LiDAR-inertial odometry system that is capable of accurately estimating the poses of both ego-vehicle and objects. We propose a trajectory-based dynamic feature filtering method, which filters out features belonging to moving objects by leveraging tracking results before scan-matching. Factor graph-based optimization is then conducted to optimize the bias of the IMU and the poses of both the ego-vehicle and surrounding objects in a sliding window. Experiments conducted on the KITTI tracking dataset and self-collected dataset show that our method achieves better pose and tracking accuracy than our previous work DL-SLOT and other baseline methods. Our open-source implementation is available at https://github.com/tiev-tongji/LIMOT.
Paper Structure (24 sections, 8 equations, 5 figures, 5 tables)

This paper contains 24 sections, 8 equations, 5 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Methods
  4. Notation and System Overview
  5. Preprocessing
  6. Object Detection
  7. IMU Pre-integration
  8. Multi-object Tracking
  9. LiDAR Odometry
  10. De-skewing, Feature Extraction and Scan-matching
  11. Trajectory-based Dynamic Feature Filtering
  12. Joint Factor Graph Optimization
  13. Experiments
  14. Evaluation Metrics
  15. Baselines
  16. ...and 9 more sections

Figures (5)

  • Figure 1: The system architecture of LIMOT. The system consists of the preprocessing, LiDAR odometry, the sliding window-based 3D multi-object tracking, and factor graph optimization.
  • Figure 2: Example of filtering out dynamic feature points. The light blue points denote the original point cloud and the red points denote the feature points. The feature points on the moving cars (b) and (c) are removed exactly by LIMOT, while the feature points on the static car (a) are remained.
  • Figure 3: Factor graph framework of LIMOT for joint optimization.
  • Figure 4: Qualitative results of LIMOT on the KITTI tracking dataset. (a) Comparison of ego-trajectories in sequence 04 of the KITTI tracking dataset. (b) Comparison of point cloud maps generated by LIO-SAM and LIMOT. (c) Trajectories of the main tracked object (id 0) and ego-vehicle in sequence 10 of the KITTI tracking dataset. (d) Comparison between the ground truth and estimated instantaneous velocity of the tracked object in (c).
  • Figure 5: The overview of the self-collected dataset. (a) LIMOT mapping result aligning with the satellite map. (b) A representative front view image of the self-collected dataset.