Autonomous Magnetometer Calibration via Extended Kalman Filter
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional magnetometer calibration methods face challenges with accuracy due to the 'hard-iron' and 'soft-iron' effects from nearby ferromagnetic components, leading to overfitting and requiring complex user gestures, especially with small or unevenly distributed measurement data sets.
Innovation Solution
An extended Kalman filter (EKF) based calibration method that utilizes sensor data from both the magnetometer and gyroscope to determine and quantify these effects in real-time, enabling autonomous calibration without user intervention, even with minimal rotational movements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional ellipsoid fitting solutions are used to calibrate magnetometers, then calibration parameters can be determined to reposition measurement points from an ellipsoid to a sphere, but overfitting occurs with small or unevenly distributed data sets which worsens calibration results
Solution Approach 1:
The system continuously monitors magnetometer measurements and compares them against the calibrated spherical model, using feedback loops to detect deviations caused by hard-iron and soft-iron effects. This ongoing validation prevents overfitting by constantly checking whether the calibration model generalizes to new measurement data rather than merely fitting the training dataset.
Solution Approach 2:
The calibration system performs self-diagnosis and self-correction by automatically detecting when measurement data becomes unevenly distributed or insufficient. It can trigger recalibration sequences, adjust data weighting, or identify when additional calibration maneuvers are needed, making the system self-aware of its own calibration quality without external intervention.
2Measurement precision
If conventional ellipsoid fitting solutions are used, then calibration can be performed, but users are prompted to perform awkward, inconvenient and complex gestures such as Figure 8 motions
Solution Approach 1:
The system dynamically adapts the calibration process based on detected device motion patterns. Instead of requiring fixed complex gestures, it can adjust calibration parameters and data weighting in real-time as the user naturally moves the device, accepting a wider variety of motion patterns while maintaining calibration quality through adaptive filtering and outlier detection.
Solution Approach 2:
The calibration system autonomously determines when sufficient calibration data has been collected by monitoring the distribution and quality of measurement points. It can prompt users for minimal additional motions only when necessary, rather than requiring predetermined complex gesture sequences, making the process self-regulating and user-adaptive.
3Measurement precision
If conventional ellipsoid fitting solutions are used, then calibration can be performed, but manual triggering is required which negatively impacts user experience
Solution Approach 1:
The system continuously monitors measurement data quality, distribution evenness, and temporal patterns to automatically determine when calibration should be performed or updated. It uses feedback from sensor fusion algorithms to detect when environmental conditions or device usage patterns indicate a need for recalibration, triggering the process without user intervention.
Solution Approach 2:
The calibration system performs self-initiated calibration by monitoring its own operational state and measurement data quality. It can automatically start calibration sequences when detecting drift, environmental changes, or insufficient data distribution, and can self-validate completion based on achieved calibration metrics, making the entire process autonomous.
4Device complexity
If magnetometers are placed near ferromagnetic components for compact device design, then device integration is improved, but hard-iron and soft-iron effects cause measurements to form an ellipsoid rather than a sphere
Solution Approach 1:
The calibration process extracts and separates the distorting effects of nearby ferromagnetic components from the true magnetic field measurements. By modeling the hard-iron and soft-iron effects as distinct mathematical components, the system can isolate and remove their influence, recovering accurate magnetic field data despite the presence of interfering ferromagnetic materials in close proximity.
Solution Approach 2:
The system transforms the measurement model from assuming a simple spherical distribution to an ellipsoidal model that accounts for ferromagnetic interference. By changing the mathematical parameters to include calibration matrices that represent the distorted measurement space, the system can compensate for the physical constraints of compact device design while maintaining measurement accuracy.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach effectively calibrates magnetometers to align measurements on a sphere, improving accuracy and reducing user inconvenience by allowing automatic, background calibration with small and uneven data sets, and maintaining optimal power usage.
Implementation Method 1
magnetized ferromagnetic components mounted on nearby printed circuit boards (PCBs) may produce a 'hard-iron effect' on a magnetometer in a handheld device
Implementation Method 2
a 'soft-iron effect' may result from the Earth's magnetic field inducing an interfering magnetic field onto normally un-magnetized ferromagnetic components of nearby PCBs
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
Systems and methods may provide for obtaining first sensor data associated with a gyroscope (22) and obtaining second sensor data associated with a magnetometer (10). Additionally, the first sensor data, the second sensor data and an extended Kalman filter may be used to calibrate the magnetometer (10). In one example, a sampling rate of the magnetometer (10) is increased before obtaining the second sensor data and the sampling rate of the magnetometer (10) is decreased after calibration of the magnetometer(10).