Magnetic Position Sensing With RNN-Based Disturbance Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing magnetic sensor systems face challenges in achieving high accuracy and robustness against external disturbances while maintaining simplicity and cost-effectiveness, particularly when determining the position of a sensor device relative to a magnetic source with varying degrees of freedom.
Innovation Solution
A method using a recurrent neural network (RNN) with a limited number of trainable parameters to determine the position of a sensor device relative to a magnetic source, utilizing a semiconductor substrate with multiple magnetic sensors and normalizing sensor signals, which is less sensitive to external disturbances and temperature variations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional magnetic sensor systems are used to determine position, then the system can measure position, but the accuracy is insufficient and the system is sensitive to external disturbances
Solution Approach 1:
The patent introduces an artificial neural network as an intermediary processing layer between the magnetic sensors and the position determination. This neural network mediator processes sensor signals and compensates for external disturbances, achieving high position accuracy (MSE < ±100 microns) while reducing sensitivity to external magnetic fields and temperature variations.
2Measurement precision
If complex algorithms are used to improve position determination accuracy, then measurement precision improves, but device complexity and computational requirements increase
Solution Approach 1:
The patent changes the approach from complex mathematical algorithms to a neural network model with optimized parameters. The neural network is trained with specific parameters (activation functions, learning rates, network architecture) to achieve high position accuracy while maintaining manageable computational complexity. The model uses standard activation functions like ReLU and sigmoid, with parameter counts kept practical for embedded systems.
3Measurement precision
If more sensor elements are added to improve measurement accuracy, then position determination improves, but device complexity and cost increase
Solution Approach 1:
The patent replaces traditional mechanical/sensor-array approaches with a neural network-based system. Instead of using multiple sensor elements arranged in complex patterns, the system uses a small number of magnetic sensors (e.g., 4-9 sensors) and processes their signals through a neural network. This substitution achieves comparable or superior accuracy with reduced hardware complexity and lower cost.
4Reliability
If the system is made more robust against positioning errors and external disturbances, then reliability improves, but device complexity increases
Solution Approach 1:
The neural network incorporates feedback mechanisms where the model continuously processes sensor signals and adjusts its internal state based on the input data. The network uses feedback from position measurements to refine its predictions, effectively compensating for positioning errors and external disturbances. This feedback approach improves reliability without requiring complex error correction hardware.
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
The method achieves high positional accuracy with a mean square error (MSE) of less than ±100 microns and is less sensitive to external disturbances, temperature, and mounting errors, while using a simple and lightweight system with reduced computational complexity.
Implementation Method 1
They are based on measuring a magnetic field characteristic at one or multiple sensor locations
Data Source
AI summary
Method of determining a position (f a sensor device movable relative to a magnetic source, or vice versa; the sensor device comprising at least two magnetic sensors; the method comprising the steps of: a) obtaining a plurality of magnetic sensor signals from said magnetic sensors; b) determining or estimating the position of the sensor device based on said plurality of sensor signals or signals derived therefrom; wherein step b) comprises: determining said position (sing an artificial neural network; the artificial neural network being a recurrent neural network trained for determining said position using at most three hundred trainable parameters per degree of freedom. A position sensor system. A position sensor device.