Smartphone Heading Correction Based on Gravity Assisted and Middle Time Simulated-Zero Velocity Update Method

Qinghua Zeng, Shijie Zeng, Jianye Liu, Qian Meng, Ruizhi Chen, Heze Huang2018 · DOI

TL;DR

This paper introduces a novel smartphone-based pedestrian navigation system that overcomes the limitations of traditional methods by using gravity-assisted and simulated zero-velocity updates. It achieves improved heading accuracy and positioning by intelligently adapting to different smartphone usage modes and compensating for sensor drift without relying on external magnetic field data.

What Problem Does This Paper Address?

Accurate pedestrian navigation, especially indoors, is a significant challenge. Traditional methods often rely on GPS, which is unreliable indoors, or inertial navigation systems (INS) that suffer from sensor drift and error accumulation. While Pedestrian Dead Reckoning (PDR) offers an autonomous solution, its accuracy is hampered by the inherent limitations of low-cost smartphone sensors (accelerometers, gyroscopes) and the unpredictable nature of smartphone usage. Specifically, the way a user holds their phone (e.g., portrait, landscape, during a call) significantly impacts sensor readings and heading calculations. Furthermore, the presence of ferromagnetic materials in urban environments can corrupt magnetic field data, making it an unreliable source for heading correction. Existing solutions often require external infrastructure (like Wi-Fi beacons) or are sensitive to environmental disturbances. This paper aims to address these issues by developing an INS-based navigation system that is robust to varying smartphone usage modes and magnetic interference.

Methods

The proposed system combines two key novel techniques:

  • Gravity Assisted (GA) Method: This method intelligently detects the smartphone's "using-mode" (normal, landscape, call) by analyzing gravity sensor data. By projecting the gravity vector onto a 3D sphere, distinct patterns emerge for each mode. This allows the system to anticipate and compensate for the significant attitude changes that occur when switching between modes, which would otherwise lead to heading errors.
  • Middle Time Simulated-Zero Velocity Update (MTS-ZUPT): Traditional Zero Velocity Update (ZUPT) algorithms rely on detecting when a foot is stationary (zero velocity). Since a smartphone is typically held, this direct detection is impossible. The MTS-ZUPT algorithm simulates this by identifying a "middle-time" within a step cycle (specifically, when one foot is in the air and the user might be changing direction). At this simulated stationary period, a Kalman Filter is used to correct the accumulated heading errors. This is an improvement over Step Time Simulated-Zero Velocity Update (STS-ZUPT), which applies the correction at the exact step time.

These methods are integrated into a Kalman Filtering framework to fuse sensor data and provide a more accurate navigation solution.

Datasets

The paper describes experiments conducted in indoor environments at the College of Automation Engineering (CAE) of NUAA. These experiments involved walking in corridors and around a rectangular area. A separate experiment was conducted outdoors on a playground using a Differential Global Positioning System (DGPS) with centimeter-level accuracy as a reference. No specific named datasets are mentioned; rather, the authors generated their own experimental data.

Key Insights and Ideas

  • Smartphone Usage Modes Matter: The orientation and way a smartphone is held significantly impact inertial sensor readings and can lead to substantial heading errors in navigation systems.
  • Gravity as a Mode Indicator: The gravity sensor, often overlooked for navigation, provides a robust way to distinguish between different smartphone usage modes, enabling proactive error compensation.
  • Simulated Zero Velocity is Viable: Even without direct zero-velocity detection, identifying a "middle-time" within a step cycle as a pseudo-stationary period allows for effective Kalman Filter-based error correction.
  • MTS-ZUPT Outperforms STS-ZUPT: Applying the simulated zero-velocity update at the "middle-time" (rather than the exact step time) yields better heading accuracy and error restraint.
  • GA Enhances Robustness: The Gravity Assisted method, when combined with MTS-ZUPT, significantly improves navigation accuracy, especially after mode changes, by compensating for heading errors introduced by attitude shifts.

Key Results

  • Step Detection Accuracy: The proposed step detection algorithm achieved an average accuracy of 99.2% in experiments.
  • MTS-ZUPT vs. STS-ZUPT:
    • In straight walking, MTS-ZUPT improved positioning accuracy by 44.8% over STS-ZUPT and 71.2% over traditional PDR.
  • In rectangular walking, MTS-ZUPT improved positioning accuracy by 72.2% over STS-ZUPT and 88.8% over traditional PDR. The walking error was reduced to approximately 0.72% of the total walking distance.
  • GA Method Effectiveness:
    • The GA algorithm successfully distinguished smartphone using-modes with 100% accuracy.
  • When integrated with MTS-ZUPT (MTS-ZUPT/GA), the system demonstrated significantly improved trajectory accuracy, especially after mode changes.
  • In an outdoor experiment with DGPS reference, the MTS-ZUPT/GA algorithm achieved a distance error of 2.0% of the walking distance (400m), with a final positioning error of (4.70, 6.92). This was superior to other tested algorithms.
  • Overall Positioning Error: The paper claims the MTS-ZUPT/GA algorithm achieves a positioning error of 1.4% to 2.0% of the walking distance in various using-mode experiments.

Limitations and Future Work

  • Sensor Accuracy: The paper acknowledges that the low-cost MEMS sensors in smartphones have poor accuracy, and body shaking can significantly impact heading. While their methods mitigate this, inherent sensor limitations remain.
  • Magnetic Field Assumption: While the method avoids relying on magnetic fields for correction, the presence of strong magnetic disturbances could still potentially affect the gravity sensor's readings, though this is not explicitly detailed as a limitation.
  • Future Work: The authors suggest that their proposed algorithm can effectively reduce using-mode errors, restrain heading divergence, and enhance smartphone navigation performance. Further validation and refinement of the algorithms in more diverse and complex environments could be future directions.

Summary updated

BibTeX
@misc{zeng2018smartphone,
  author = {Qinghua Zeng and Shijie Zeng and Jianye Liu and Qian Meng and Ruizhi Chen and Heze Huang},
  title = {{Smartphone Heading Correction Based on Gravity Assisted and Middle Time Simulated-Zero Velocity Update Method}},
  year = {2018},
  doi = {10.3390/s18103349},
}
Figure