TL;DR
This paper introduces a novel geomagnetic fingerprinting method, Ellipse Coefficient Map (ECM)-based UFEE, that significantly improves indoor positioning accuracy by accounting for azimuth angle variations. By representing 2D magnetic vectors as ellipse coefficients and applying a weighted equalization method for the Z-axis, the system achieves higher accuracy and a smaller database size compared to traditional methods.
What Problem Does This Paper Address?
Indoor positioning systems are crucial for navigation in environments where GPS is unavailable. Geomagnetic fingerprinting is a promising technique because it leverages the Earth's magnetic field, which is relatively stable over time and doesn't require additional infrastructure like Wi-Fi access points. However, a major challenge with geomagnetic fingerprinting is that the measured magnetic signals at a single location can vary significantly with changes in the device's orientation (azimuth angle). This necessitates large databases that store signal strengths for numerous angles, or it leads to substantial positioning errors if only a few angles are recorded. Existing methods often struggle to accurately estimate positions when the device's orientation deviates from the trained angles, leading to a trade-off between database size and positioning robustness.
Methods
The proposed method, Ellipse Coefficient Map (ECM)-based UFEE, addresses the azimuth angle problem by:
- Ellipse Coefficient Map (ECM): Instead of storing raw magnetic signal magnitudes, the system captures the 2D characteristics of magnetic vectors (X and Y axes) at each reference point (RP) by fitting an ellipse. The ellipse coefficients (a, b, c, d, e, f) define the shape, orientation, and size of the ellipse formed by magnetic vectors across different azimuth angles. This significantly reduces the database size.
- Normalized Linear Least Squares: To accurately derive these ellipse coefficients and minimize curvature bias, the paper employs a normalized linear least squares method. This method is applied to at least five 2D magnetic vector measurements taken at arbitrary angles.
- Weighted Geomagnetic Signal Equalization: The Z-axis geomagnetic signal is often much larger than the X and Y components and can disproportionately affect positioning accuracy. To mitigate this, a weighted geomagnetic signal equalization method is used. This method calculates an optimal "Ratio" to scale the Z-axis values, ensuring they have a comparable influence to the ellipse coefficients during the positioning phase.
- Positioning Phase: In the positioning phase, real-time magnetic vector measurements are used to calculate an ellipse equation. The system then compares these real-time ellipse coefficients and the compensated Z-axis values with the stored ECMs at each RP. The position with the minimum distance (calculated using a specific formula involving ellipse coefficients and compensated Z-values) is determined as the user's location.
Datasets
The experiments were conducted in indoor common corridors and rooms at the Korea Maritime and Ocean University. The paper describes collecting data from 23 reference points per floor across three floors (2nd, 3rd, and 4th), with a 3m reference point interval. For training the ECM, 15 data points were collected at 0, 70, 140, 210, and 280° azimuths at each RP. An additional 15 data points were used for positioning at each RP. The paper does not explicitly mention using publicly available datasets but rather describes data collected from a specific experimental environment.
Key Insights and Ideas
- Representing Magnetic Signatures as Ellipses: The core innovation is using ellipse coefficients to represent the 2D magnetic field characteristics at a location. This is a more robust representation than raw signal magnitudes, as it captures the inherent elliptical nature of magnetic vectors due to sensor and environmental effects.
- Reducing Database Size: By storing ellipse coefficients instead of raw data for multiple angles, the database size is dramatically reduced, making the system more practical.
- Handling Z-axis Dominance: The weighted equalization of the Z-axis signal is crucial. It prevents the large Z-axis values from overwhelming the positioning calculation, allowing the ellipse coefficients to contribute more effectively.
- Azimuth Invariance: The ECM approach inherently handles azimuth variations by characterizing the entire elliptical pattern, rather than relying on specific angle measurements.
Key Results
The proposed ECM-based UFEE method demonstrated significant improvements over the existing geomagnetic intensity-based fingerprinting method (MGIF [10]):
- Positioning Accuracy: The average positioning accuracy for the proposed UFEE was 89.14% within a 1.5m radius, compared to 27.00% for MGIF.
- Distance Error: The average distance error was reduced from 1.00m (MGIF) to 0.13m (UFEE) for the minimum error, and from 12.00m to 4.53m for the maximum error across the tested floors.
- Database Size: The paper implies a significant reduction in database size due to storing ellipse coefficients rather than raw data for numerous angles.
Limitations and Future Work
- Environmental Dependence: The paper notes that the "Ratio" for the weighted geomagnetic signal equalization method is specific to the applied indoor environment and needs to be derived for each new environment.
- Initial Position Estimation: The paper mentions that geomagnetic fingerprinting, like other fingerprinting methods, can face challenges in estimating the initial start position without fixed APs.
- Future Work: While not explicitly detailed, the success of this method suggests future work could involve exploring more sophisticated equalization techniques, optimizing the number of angles needed for ECM generation, and testing the system in a wider variety of indoor environments. The paper also mentions the potential for integrating this method with other indoor positioning technologies.
Summary updated