An Improved Heuristic Drift Elimination Method for Indoor Pedestrian Positioning

Zhi-Hong Deng, Yun Cao, Pengyu Wang, Bo Wang2018 · DOI

TL;DR

This paper introduces an improved Heuristic Drift Elimination (HDE) method (IHDE) for indoor pedestrian positioning. By intelligently evaluating pedestrian motion and considering dominant building directions, IHDE significantly reduces heading drift errors, outperforming existing methods in complex indoor environments.

What Problem Does This Paper Address?

Indoor positioning is a critical area with growing demand for location-based services. Pedestrian Dead Reckoning (PDR) is a popular technique for indoor positioning, relying on inertial sensors. However, PDR suffers from accumulating errors, particularly heading drift, which degrades accuracy over time. While the Heuristic Drift Elimination (HDE) algorithm offers a way to correct heading drift by leveraging dominant building directions (like walls), it struggles when pedestrians move along non-dominant paths or in non-straight lines, leading to overcorrection and new errors. This paper aims to address this limitation by proposing an improved HDE method that is more robust to varied indoor movement patterns.

Methods

The proposed method, Improved Heuristic Drift Elimination (IHDE), builds upon the standard IEZ (Inertial Navigation System, Extended Kalman Filter, Zero Velocity Update) framework. The core novelty lies in a multi-stage approach to heading correction:

  • Motion Comprehension: The system first analyzes the pedestrian's motion. It uses a Z-type membership function based on the standard deviation of heading angles over a short window to determine if the pedestrian is moving in a straight line. Additionally, a bell-shaped membership function assesses the uniformity of step intervals to detect abnormal motion states.
  • Comprehensive Evaluation: These two metrics are combined using a weighted sum to create a comprehensive evaluation score. This score determines if the IHDE method should be triggered. The method is only activated if the pedestrian is moving in a relatively straight line and not exhibiting abnormal motion.
  • Dominant Direction Check: If the IHDE method is triggered, it then checks if the pedestrian is moving along dominant building directions (defined as 0°, 90°, 180°, 270°).
  • Heading Error Measurement: If the path is along a dominant direction, a heading error is calculated and fed into the EKF for correction. If the path is not along a dominant direction but is still deemed a straight line, a different reference direction is used for error calculation. Crucially, if the motion is not a straight line (e.g., a circular path), the method avoids heading correction to prevent overcorrection.

Datasets

The paper does not explicitly mention or introduce new datasets. Instead, it relies on experimental tests conducted in various indoor environments (an IT laboratory building, an outdoor playground) to evaluate the proposed method.

Key Insights and Ideas

  • Motion-Aware Heading Correction: The most significant insight is the development of a motion comprehension module that intelligently decides when to apply heading correction. This moves beyond simply assuming dominant directions are always relevant.
  • Avoiding Overcorrection: By explicitly identifying non-straight paths and abnormal motion, the IHDE method effectively prevents the overcorrection issues that plague the original HDE algorithm in complex environments.
  • Adaptability to Environment: The method's ability to adapt to different movement patterns (straight lines, curves, dominant and non-dominant directions) is key to its improved performance.
  • Integration with IEZ: The IHDE method is seamlessly integrated into the existing IEZ framework, leveraging its established components for motion estimation.

Key Results

The paper demonstrates the effectiveness of IHDE through several tests:

  • Test 1 (Dominant Directions): In a path composed of straight lines and right angles, IHDE achieved a positioning error of 0.2% to 0.4%, significantly outperforming IEZ (8.7%–9.6%) and HDE (0.3%–0.6%).
  • Test 2 (Mixed Directions): In a path with dominant, non-dominant, and curved sections, IHDE maintained a low error (0.4%–0.7%), while HDE struggled with non-dominant straight segments, leading to a larger error (1.3%–1.7%). IEZ also showed significant drift.
  • Test 3 (Curved Paths): For paths involving curved sections, IHDE correctly avoided heading correction, resulting in a small deviation (approx. 2m error). In contrast, HDE's continuous correction based on dominant directions rendered the trajectory useless.

Overall, IHDE consistently achieved the lowest positioning errors across various scenarios.

Limitations and Future Work

  • Limitations:
    • The paper acknowledges that the original HDE algorithm is not suitable for non-dominant direction conditions, which IHDE addresses. However, the performance in highly complex, non-linear indoor environments (e.g., highly cluttered spaces with frequent sharp turns) might still be a challenge.
  • The method relies on the accuracy of the underlying IEZ framework, including the Zero Velocity Detector (ZVD). Errors in ZVD can propagate.
  • Future Work:
    • The authors suggest that the proposed method demonstrates adaptability to different environments. Further testing in more diverse and complex indoor settings would be beneficial.
  • While the paper focuses on heading correction, exploring further enhancements to other aspects of PDR for even greater accuracy could be a future direction.

Summary updated

BibTeX
@misc{deng2018improved,
  author = {Zhi-Hong Deng and Yun Cao and Pengyu Wang and Bo Wang},
  title = {{An Improved Heuristic Drift Elimination Method for Indoor Pedestrian Positioning}},
  year = {2018},
  doi = {10.3390/s18061874},
}
Figure