Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Robust Vehicle Localization and Tracking in Rain using Street Maps

Yu Xiang Tan, Malika Meghjani

TL;DR

Map-Fusion is a sensor-fusion framework that leverages 2D street-network maps to correct drifting VO/VIO and intermittent GPS in adverse urban scenarios. By converting OpenStreetMap data into a road graph and matching pose estimates to the map within a factor-graph, it applies road-alignment priors to maintain localization when visual inputs are unreliable due to rain or tunnels. The approach demonstrates consistent improvements in Absolute Trajectory Error across KITTI, Oxford Robotcar, 4Seasons, and Singapore datasets, including successful real-time demonstrations on a hardware-constrained mobile robot. While effective, its performance depends on map accuracy and has limited capability to correct motion-direction drifts, motivating future work to incorporate additional sensing modalities such as Radar or LiDAR. Overall, Map-Fusion provides a practical, scalable enhancement to existing VO/VIO systems for urban vehicle tracking under challenging weather conditions.

Abstract

GPS-based vehicle localization and tracking suffers from unstable positional information commonly experienced in tunnel segments and in dense urban areas. Also, both Visual Odometry (VO) and Visual Inertial Odometry (VIO) are susceptible to adverse weather conditions that causes occlusions or blur on the visual input. In this paper, we propose a novel approach for vehicle localization that uses street network based map information to correct drifting odometry estimates and intermittent GPS measurements especially, in adversarial scenarios such as driving in rain and tunnels. Specifically, our approach is a flexible fusion algorithm that integrates intermittent GPS, drifting IMU and VO estimates together with 2D map information for robust vehicle localization and tracking. We refer to our approach as Map-Fusion. We robustly evaluate our proposed approach on four geographically diverse datasets from different countries ranging across clear and rain weather conditions. These datasets also include challenging visual segments in tunnels and underpasses. We show that with the integration of the map information, our Map-Fusion algorithm reduces the error of the state-of-the-art VO and VIO approaches across all datasets. We also validate our proposed algorithm in a real-world environment and in real-time on a hardware constrained mobile robot. Map-Fusion achieved 2.46m error in clear weather and 6.05m error in rain weather for a 150m route.

Robust Vehicle Localization and Tracking in Rain using Street Maps

TL;DR

Map-Fusion is a sensor-fusion framework that leverages 2D street-network maps to correct drifting VO/VIO and intermittent GPS in adverse urban scenarios. By converting OpenStreetMap data into a road graph and matching pose estimates to the map within a factor-graph, it applies road-alignment priors to maintain localization when visual inputs are unreliable due to rain or tunnels. The approach demonstrates consistent improvements in Absolute Trajectory Error across KITTI, Oxford Robotcar, 4Seasons, and Singapore datasets, including successful real-time demonstrations on a hardware-constrained mobile robot. While effective, its performance depends on map accuracy and has limited capability to correct motion-direction drifts, motivating future work to incorporate additional sensing modalities such as Radar or LiDAR. Overall, Map-Fusion provides a practical, scalable enhancement to existing VO/VIO systems for urban vehicle tracking under challenging weather conditions.

Abstract

GPS-based vehicle localization and tracking suffers from unstable positional information commonly experienced in tunnel segments and in dense urban areas. Also, both Visual Odometry (VO) and Visual Inertial Odometry (VIO) are susceptible to adverse weather conditions that causes occlusions or blur on the visual input. In this paper, we propose a novel approach for vehicle localization that uses street network based map information to correct drifting odometry estimates and intermittent GPS measurements especially, in adversarial scenarios such as driving in rain and tunnels. Specifically, our approach is a flexible fusion algorithm that integrates intermittent GPS, drifting IMU and VO estimates together with 2D map information for robust vehicle localization and tracking. We refer to our approach as Map-Fusion. We robustly evaluate our proposed approach on four geographically diverse datasets from different countries ranging across clear and rain weather conditions. These datasets also include challenging visual segments in tunnels and underpasses. We show that with the integration of the map information, our Map-Fusion algorithm reduces the error of the state-of-the-art VO and VIO approaches across all datasets. We also validate our proposed algorithm in a real-world environment and in real-time on a hardware constrained mobile robot. Map-Fusion achieved 2.46m error in clear weather and 6.05m error in rain weather for a 150m route.
Paper Structure (16 sections, 9 equations, 6 figures, 4 tables)

This paper contains 16 sections, 9 equations, 6 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Map-based Localization
  4. Sensor Fusion Approaches
  5. Localization in Adverse Weather
  6. Approach
  7. Processing Map Information
  8. Matching Vehicle Pose Estimate to Map
  9. GPS Initialization and Scaling
  10. Experiments
  11. Datasets and Evaluation
  12. Experimental Setup
  13. Quantitative Results
  14. Qualitative Results
  15. Real-World Results
  16. ...and 1 more sections

Figures (6)

  • Figure 1: Map-Fusion: Correcting ORB-SLAM3 VIO odometry estimates for a rain + tunnel sequence in Munich from the 4Seasons Dataset. Without Map-Fusion, the odometry drifts away from lane into a different part of the city.
  • Figure 2: Overview of Map-Fusion.
  • Figure 3: Sample images from the four datasets.
  • Figure 4: Output trajectories of DROID-SLAM (left) and DROID-SLAM + Map-Fusion(right) for Oxford Robotcar sequences 2014-11-25-09-18-32 (top), 2015-10-29-12-18-17 (middle), 2014-12-09-13-21-02 (bottom)
  • Figure 5: Output trajectories of ORB-SLAM3 (left) and ORB-SLAM3 + Map-Fusion (right) for 4Seasons sequence city_loop_1_train
  • ...and 1 more figures