Robust Anchor-Aided GNSS/PDR Pedestrian Localization via Factor Graph Optimization for Remote Sighted Assistance

Sen Huang, Jinjing Zhao, Yihan Zhong, Yiding Liu, Shengyong XuSensors 2025 · DOI

TL;DR

This paper introduces a robust pedestrian localization system for visually impaired individuals using a combination of GNSS, PDR, and novel "road anchor" constraints within a factor graph optimization framework. This approach significantly improves navigation accuracy in challenging urban environments, enabling the creation of video-based maps for enhanced remote sighted assistance.

What Problem Does This Paper Address?

Remote Sighted Assistance (RSA) systems aim to help visually impaired people (VIPs) navigate by connecting them with sighted agents who provide guidance. However, unfamiliar urban environments can be disorienting for both VIPs and agents, leading to anxiety and reduced effectiveness of the assistance. A key challenge is providing accurate and reliable real-time location information for the VIP, especially in areas where GNSS signals are weak or unreliable (e.g., urban canyons). Existing methods often struggle with drift and accuracy in these conditions. This paper addresses the need for a more precise and robust localization system that can support advanced RSA features like video-based map previews.

Methods

The core of the proposed solution is a Factor Graph Optimization (FGO) framework that fuses data from multiple sources:

  • Global Navigation Satellite System (GNSS): Provides absolute global position references, acting as a global constraint.
  • Pedestrian Dead Reckoning (PDR): Uses smartphone inertial sensors (accelerometer, gyroscope, magnetometer) to estimate relative motion (step detection, step length, heading). PDR provides continuous, high-frequency motion updates but suffers from cumulative drift.
  • Novel Road Anchor Constraints: This is the key innovation. The system identifies "turning points" in the environment (e.g., corners of buildings, intersections) and uses their known geographic coordinates as anchors. A turning-point-based anchor-matching method aligns the estimated trajectory with these anchors. These anchors provide crucial spatial constraints, especially in areas where PDR is prone to drift.

The FGO framework employs a coarse-to-fine optimization strategy. First, GNSS, PDR, and initial start/end point constraints are used for a coarse trajectory estimation. Then, the anchor-matching algorithm identifies the best matches between the coarse trajectory and predefined road anchors. Finally, these anchor constraints are incorporated for a fine-grained optimization, yielding a more accurate trajectory.

Datasets

The authors collected data in three representative urban environments characterized by dense buildings and vegetation, which typically degrade GNSS performance. They used a consumer-grade smartphone (OPPO PESM10) for data acquisition. Ground truth trajectories were obtained using a Vision-RTK 2 system, providing centimeter-level accuracy at 10 Hz. Experiments were conducted with multiple datasets from different walking routes and even with a different user to assess robustness. No new benchmarks were introduced; rather, existing urban environments were used to test the proposed method against established baselines.

Key Insights and Ideas

  • Anchors are Crucial in Urban Canyons: The most significant insight is the effectiveness of "road anchors" (turning points) in correcting accumulated drift, particularly in complex urban environments where GNSS is unreliable and PDR naturally drifts.
  • Coarse-to-Fine is Key: The two-stage optimization (coarse estimation followed by anchor-based fine-tuning) is essential for efficiently leveraging both global (GNSS) and local (anchor) constraints.
  • Video-Based Maps are Feasible: The improved localization accuracy directly enables the creation of detailed video-based maps, which are valuable for RSA by providing agents and VIPs with a preview of the environment.
  • Consumer-Grade Hardware is Sufficient: The method demonstrates strong performance even with a standard consumer smartphone, suggesting broad applicability.

Key Results

The proposed anchor-aided FGO-GNSS/PDR method significantly outperforms GNSS-only and FGO-GNSS/PDR baselines.

  • Mean Horizontal Positioning Error: Reduced by 42% to 65% across all datasets. The overall mean error achieved was 1.36 m.
  • Maximum Error: Reduced by 38% to 76% across all datasets.
  • Specific Improvements: In Experiment 1, the mean error was reduced from 2.39m (FGO-GNSS/PDR) to 0.97m (FGO-GNSS/PDR + Anchor). In Experiment 2, the mean error dropped from 3.79m to 1.63m (a 57% improvement). In Experiment 3, the mean error was 1.19m.
  • Robustness: The method showed consistent performance across different datasets and even with a different user, indicating good robustness.

Limitations and Future Work

  • Anchor Matching Limitations: The authors acknowledge that the anchor-matching algorithm still has certain limitations, and in some scenarios, relatively large matching errors can persist.
  • Generalizability: While tested in diverse urban settings, further evaluation across more varied routes and scenarios is needed to fully validate generalization and robustness across different users and walking patterns.
  • Future Work:
    • Incorporate raw GNSS measurements for potentially higher accuracy.
  • Enhance the PDR algorithm with improved step detection and heading estimation.
  • Precisely define anchor positions and types, and develop a more robust anchor-matching algorithm.
  • Implement a real-time, sliding window-based FGO framework for real-time pedestrian localization in RSA applications.

Summary updated

Abstract

Remote Sighted Assistance (RSA) systems provide visually impaired people (VIPs) with real-time guidance by connecting them with remote sighted agents to facilitate daily travel. However, unfamiliar environments often complicate decision-making for agents and can induce anxiety in VIPs, thereby reducing the effectiveness of the assistance provided. To address this challenge, this paper proposes a video-based map assistance method. By pre-recording pedestrian path videos and aligning them with geographic locations, the system enables route preview and enhances navigation guidance. This study introduces a factor graph optimization (FGO) algorithm that integrates Global Navigation Satellite System (GNSS) and pedestrian dead reckoning (PDR) data for pedestrian positioning. It incorporates road-anchor constraints, a turning-point-based anchor-matching method, and a coarse-to-fine optimization strategy to improve the positioning accuracy. GNSS provides global reference positions, PDR offers precise relative motion constraints through accurate heading estimation, and anchor factors further enhance localization accuracy by leveraging known geometric features. We collected data using a smartphone equipped with a four-camera module and conducted tests in representative urban environments. Experimental results demonstrate that the proposed anchor-aided FGO-GNSS/PDR algorithm achieves robust and accurate positioning, effectively supporting video-based map construction in complex urban settings. With anchor constraints, the mean horizontal positioning error was reduced by 42% to 65% and the maximum error by 38% to 76% across all datasets. In this study, the mean horizontal positioning error was 1.36 m.
BibTeX
@inproceedings{huang2025robust,
  author = {Sen Huang and Jinjing Zhao and Yihan Zhong and Yiding Liu and Shengyong Xu},
  title = {{Robust Anchor-Aided GNSS/PDR Pedestrian Localization via Factor Graph Optimization for Remote Sighted Assistance}},
  booktitle = {Sensors},
  year = {2025},
  doi = {10.3390/s25175536},
}
Figure