Sensor Fusion Algorithm Reducing Computational Complexity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sensor fusion algorithms, such as the Kalman Filter, face high computational complexity and power consumption when integrating data from accelerometers, gyroscopes, and magnetometers, making it necessary to implement them on an external microprocessor, which increases chip area and power requirements.
Innovation Solution
A comprehensive sensor fusion algorithm using a Factored Quaternion Algorithm (FQA) to combine acceleration rates and magnetic field magnitudes, dynamically merged with a gyro output through a complementary filter, reducing computational complexity and enabling local integration within the sensor system.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a Kalman Filter is applied to integrate sensor data from accelerometers, gyroscopes and magnetometers, then measurement precision is improved, but device complexity and power consumption increase significantly
Solution Approach 1:
The patent segments the sensor fusion problem into distinct processing stages: accelerometer data processing, gyroscope data processing, and magnetometer data processing, with each sensor type handled by specialized processing modules rather than a monolithic Kalman Filter. This segmentation reduces computational complexity while maintaining measurement precision through targeted processing of each sensor's unique characteristics
Solution Approach 2:
The patent changes the algorithmic parameters from a full-state Kalman Filter (7 states, 100 MIPS) to optimized processing algorithms that reduce computational requirements. By adjusting the processing parameters and state representations, the system achieves comparable measurement precision with significantly reduced computational complexity suitable for embedded implementation
2Measurement precision
If a Kalman Filter is applied to integrate sensor data, then measurement precision is improved, but power consumption increases
Solution Approach 1:
The patent divides the power-consuming sensor fusion operation into separate, optimized processing paths for each sensor type. This segmentation allows each processing module to use algorithms tailored to its specific computational requirements, reducing overall power consumption while maintaining the precision benefits of multi-sensor integration
Solution Approach 2:
The patent modifies the computational parameters and processing frequency of the sensor fusion algorithm to reduce power consumption. By optimizing the processing parameters and using more efficient mathematical operations, the system maintains measurement precision while reducing the energy required to sustain normal operation
3Measurement precision
If sensor fusion is implemented on an external microprocessor, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent merges the sensor fusion processing functionality directly into the sensor system's integrated circuit, combining the processing logic with the sensor arrays. This integration eliminates the need for external microprocessor implementation, reducing chip area and device complexity while maintaining measurement precision through on-chip processing
Solution Approach 2:
The patent implements a nested architecture where the processing modules are integrated within the sensor system's internal structure. The processing logic is embedded within the sensor array's operational framework, creating a compact, self-contained system that reduces overall device complexity while preserving the precision benefits of sensor fusion
Data Source
AI summary
The present invention relates to a signal processor, and more particularly, to systems, devices and methods of using a comprehensive sensor fusion algorithm to integrate sensor data collected by accelerometers, gyroscopes and magnetometers. The signal processor dynamically applies a Complementary Filter to merge a rotation based gyro output and a FQA output that is obtained by combining acceleration rates and magnetic field magnitudes. Therefore, motion information is derived to generate a motion control signal. Such a comprehensive sensor fusion algorithm significantly reduces the complexity of computation, and the power and area overhead is controlled. As a result, the signal processor may be implemented based on local computation capability of a sensor system, and its integration within such a sensor system is made possible without relying on an external microprocessor.