An Algorithm with Iteration Uncertainty Eliminate Based on Geomagnetic Fingerprint under Mobile Edge Computing for Indoor Localization

Jie Li, Liming Sun, Dongpeng Liu, Ruiyun Yu, Xingwei Wang2022 · DOI

TL;DR

This paper introduces GeoLoc, a novel indoor localization algorithm that leverages geomagnetic fingerprinting and inertial sensor data to achieve high accuracy without relying on external infrastructure. By employing an iterative uncertainty elimination approach and mobile edge computing, GeoLoc addresses the limitations of existing methods, offering a robust and efficient solution for indoor location-based services.

What Problem Does This Paper Address?

Accurate indoor localization is a persistent challenge. Traditional methods like GPS are ineffective indoors due to signal obstruction. Existing indoor positioning systems (IPS) often rely on infrastructure like Wi-Fi access points or beacons, which can be costly to deploy and maintain. While magnetic fingerprinting offers an infrastructure-free alternative, it faces issues like the non-uniqueness of magnetic fingerprints and sensor measurement deviations. Furthermore, many existing methods struggle with "cold start" problems (initial inaccuracy) and error accumulation over time, especially on resource-constrained mobile devices. This paper aims to overcome these limitations by developing a precise, infrastructure-free, and computationally efficient indoor localization system.

Methods

GeoLoc combines several key techniques:

  • Geomagnetic Fingerprinting: The system builds a magnetic map of the indoor environment. It then matches the real-time magnetic sensor readings from a smartphone against this map to estimate the user's position.
  • Multi-Sensor Fusion: GeoLoc fuses data from the smartphone's accelerometer and gyroscope with magnetic field data. This fusion helps to refine orientation estimation and mitigate errors.
  • Iterative Uncertainty Elimination: This is a core contribution. The algorithm iteratively refines position estimates by gradually reducing uncertainty. It starts with a broad set of candidate positions and progressively narrows them down based on new sensor data and motion models. This approach helps to address both cold start issues and error accumulation.
  • Pedestrian Dead Reckoning (PDR): GeoLoc incorporates a PDR model, enhanced with a dynamic motion model, to estimate step length and direction, further improving trajectory tracking.
  • Mobile Edge Computing (MEC): The algorithm is designed to be deployed on edge devices, reducing latency and computational load on a central server, making it suitable for real-time applications on smartphones.
  • Kalman Filter: A Kalman filter is used to smooth the trajectory and fuse the various sensor inputs and estimations.

Datasets

The authors collected magnetic signals on the fifth and sixth floors of an information building at Northeastern University in China. They created a 40x40 cm grid for fingerprinting and collected data multiple times over two months to ensure stability. The data was collected using Samsung Note 4 and Meizu Note 6 phones.

Key Insights and Ideas

  • Geomagnetic Field as a Stable Indoor Fingerprint: The paper highlights the stability and uniqueness of the geomagnetic field in indoor environments, making it a promising candidate for infrastructure-free localization.
  • Iterative Uncertainty Reduction is Key: The core innovation lies in the iterative process of reducing uncertainty. By starting with a broad set of possibilities and progressively refining them, the algorithm effectively handles initial inaccuracies and cumulative errors.
  • Mobile Edge Computing for Real-Time Performance: Deploying the algorithm on edge devices is crucial for achieving the low latency required for real-time indoor localization.
  • Fusion of Inertial and Magnetic Data: Combining data from accelerometers, gyroscopes, and magnetometers provides a more robust estimation than any single sensor alone.

Key Results

  • Accuracy: The algorithm achieves an accuracy of less than 2.5 m in indoor environments.
  • Robustness: GeoLoc demonstrates good performance even with variations in user trajectories, including turns.
  • Infrastructure-Free: The system does not require any external infrastructure like beacons or Wi-Fi access points.
  • Cold Start and Error Accumulation: The iterative uncertainty elimination approach significantly mitigates cold start problems and reduces error accumulation compared to traditional PDR or Kalman filter methods.
  • Efficiency: The use of mobile edge computing and a refined algorithm leads to efficient performance on smartphones.

Limitations and Future Work

  • Fingerprint Map Creation: Building the geomagnetic fingerprint map is described as time-consuming.
  • Sensor Data Dependency: The system relies on the availability and quality of smartphone sensors.
  • Generalization: While tested in one building, the authors suggest future work to evaluate the model in a larger number of buildings to identify potential parametric patterns.
  • Integration of More Sensors: Future work could involve integrating more built-in smartphone sensors to further improve accuracy.
  • SLAM/Crowdsourcing: The authors propose using Simultaneous Localization and Mapping (SLAM) frameworks or crowdsourcing methods to construct geomagnetic fingerprint maps more efficiently.

Summary updated

BibTeX
@misc{li2022algorithm,
  author = {Jie Li and Liming Sun and Dongpeng Liu and Ruiyun Yu and Xingwei Wang},
  title = {{An Algorithm with Iteration Uncertainty Eliminate Based on Geomagnetic Fingerprint under Mobile Edge Computing for Indoor Localization}},
  year = {2022},
  doi = {10.3390/s22239032},
}
Figure