Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Localization Meets Uncertainty: Uncertainty-Aware Multi-Modal Localization

Hye-Min Won, Jieun Lee, Jiyong Oh

TL;DR

This work tackles robust indoor localization by introducing an uncertainty-aware end-to-end localization framework that does not modify the underlying prediction model. It extends FusionLoc to output aleatoric and epistemic uncertainties and applies a percentile-based rejection strategy to discard high-uncertainty predictions, thereby improving accuracy in multi-modal RGB and 2D LiDAR localization. Experimental results across three real-world datasets show consistent reductions in both position and orientation errors as uncertainty thresholds become stricter, with notable improvements on noisier environments. The approach offers a practical, outlier-robust mechanism to increase reliability for global localization and can serve as a precursor to traditional localization modules in real deployments.

Abstract

Reliable localization is critical for robot navigation in complex indoor environments. In this paper, we propose an uncertainty-aware localization method that enhances the reliability of localization outputs without modifying the prediction model itself. This study introduces a percentile-based rejection strategy that filters out unreliable 3-DoF pose predictions based on aleatoric and epistemic uncertainties the network estimates. We apply this approach to a multi-modal end-to-end localization that fuses RGB images and 2D LiDAR data, and we evaluate it across three real-world datasets collected using a commercialized serving robot. Experimental results show that applying stricter uncertainty thresholds consistently improves pose accuracy. Specifically, the mean position error is reduced by 41.0%, 56.7%, and 69.4%, and the mean orientation error by 55.6%, 65.7%, and 73.3%, when applying 90%, 80%, and 70% thresholds, respectively. Furthermore, the rejection strategy effectively removes extreme outliers, resulting in better alignment with ground truth trajectories. To the best of our knowledge, this is the first study to quantitatively demonstrate the benefits of percentile-based uncertainty rejection in multi-modal end-to-end localization tasks. Our approach provides a practical means to enhance the reliability and accuracy of localization systems in real-world deployments.

Localization Meets Uncertainty: Uncertainty-Aware Multi-Modal Localization

TL;DR

This work tackles robust indoor localization by introducing an uncertainty-aware end-to-end localization framework that does not modify the underlying prediction model. It extends FusionLoc to output aleatoric and epistemic uncertainties and applies a percentile-based rejection strategy to discard high-uncertainty predictions, thereby improving accuracy in multi-modal RGB and 2D LiDAR localization. Experimental results across three real-world datasets show consistent reductions in both position and orientation errors as uncertainty thresholds become stricter, with notable improvements on noisier environments. The approach offers a practical, outlier-robust mechanism to increase reliability for global localization and can serve as a precursor to traditional localization modules in real deployments.

Abstract

Reliable localization is critical for robot navigation in complex indoor environments. In this paper, we propose an uncertainty-aware localization method that enhances the reliability of localization outputs without modifying the prediction model itself. This study introduces a percentile-based rejection strategy that filters out unreliable 3-DoF pose predictions based on aleatoric and epistemic uncertainties the network estimates. We apply this approach to a multi-modal end-to-end localization that fuses RGB images and 2D LiDAR data, and we evaluate it across three real-world datasets collected using a commercialized serving robot. Experimental results show that applying stricter uncertainty thresholds consistently improves pose accuracy. Specifically, the mean position error is reduced by 41.0%, 56.7%, and 69.4%, and the mean orientation error by 55.6%, 65.7%, and 73.3%, when applying 90%, 80%, and 70% thresholds, respectively. Furthermore, the rejection strategy effectively removes extreme outliers, resulting in better alignment with ground truth trajectories. To the best of our knowledge, this is the first study to quantitatively demonstrate the benefits of percentile-based uncertainty rejection in multi-modal end-to-end localization tasks. Our approach provides a practical means to enhance the reliability and accuracy of localization systems in real-world deployments.

Paper Structure

This paper contains 12 sections, 7 equations, 6 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. End-to-end localization
  4. Uncertainty quantification
  5. Method
  6. FusionLoc
  7. Measureing uncertainty
  8. Uncertainty-aware localization
  9. Experiments
  10. Datasets
  11. Evaluation
  12. Conclusions

Figures (6)

  • Figure 1: Our pipeline for uncertainty-aware end-to-end localization
  • Figure 2: A serving robot used in this study: The blue box represents a SLAMTEC RPLiDAR A1M8, while the red boxes indicate Intel RealSense D435 cameras.
  • Figure 3: Robot trajectories in each dataset.
  • Figure 4: Visualization of robot trajectories in different scenarios. (a) Full-loop trajectory. (b) Zigzag navigation. (c) Localized back-and-forth motion. (d) In-place rotations at specific locations.
  • Figure 5: Trajectory visualization with different uncertainty thresholds (100%, 90%, 80%, 70%). (a) TheGardenParty. (b) ETRI: Full-loop trajectory. (c) ETRI: Zigzag navigation. (d) ETRI: Localized back-and-forth motion. (e) ETRI: In-place rotations at specific locations. (f) SusungHotel.
  • ...and 1 more figures