Particle Filter Sensor Fusion for Position Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional PNT systems, such as GPS and IMUs, face limitations in accuracy and reliability due to low update rates and susceptibility to signal loss, especially in non-linear and non-Gaussian environments, which complicates navigation in environments like airborne imaging applications.
Innovation Solution
The implementation of a particle filter-based data fusion technique that up-samples position estimates by assigning sensor weights to each particle filter output, allowing for more accurate and rapid position estimation by prioritizing sensor measurements, thereby enhancing navigation solutions in complex scenarios.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If GPS is used to determine location, then position accuracy is improved (approximately 30 cm with L5 band), but update rate deteriorates (low frequency typically on the order of 50 Hz)
Solution Approach 1:
The patent combines GPS measurements with IMU measurements into a unified navigation solution using a particle filter. The GPS provides accurate position updates while the IMU provides high-rate motion data, and the particle filter fuses these disparate data sources to produce both accurate and high-rate position estimates.
Solution Approach 2:
The particle filter acts as an intermediary that processes both GPS and IMU measurements. It uses the GPS measurements to correct IMU drift while maintaining the high update rate capability of the IMU, effectively mediating between the conflicting requirements of accuracy and update rate.
2Speed
If IMU is used to provide position estimates, then update rate is improved (100 to 200 Hz), but measurement precision deteriorates (IMU drift and compounding errors)
Solution Approach 1:
The particle filter implements feedback by using GPS measurements to continuously correct the IMU-based position estimates. When GPS data is available, it provides feedback to reduce the accumulated drift errors in the IMU integration, maintaining long-term accuracy while preserving the high update rate.
Solution Approach 2:
The patent merges the high-rate IMU data with periodic GPS corrections in a unified particle filter framework. This combination allows the system to maintain the fast update rate of the IMU while periodically correcting drift using the more accurate but slower GPS measurements.
3Device complexity
If traditional Kalman filter is used for sensor fusion, then computational simplicity is improved, but adaptability deteriorates (limited to linear systems with additive, white Gaussian noise)
Solution Approach 1:
The patent changes the fundamental parameters and assumptions of the filtering approach by using a particle filter instead of a Kalman filter. This allows the system to handle non-linear dynamics and non-Gaussian noise characteristics that are common in airborne imaging applications, while still maintaining computational tractability through efficient particle filter implementations.
Data Source
AI summary
A method comprises: obtaining a GPS measurement; obtaining a first IMU measurement; obtaining a second IMU measurement; applying a first particle filter to the GPS measurement and the first IMU measurement to obtain a first position solution; applying a second particle filter to the GPS measurement and the second IMU measurement to obtain a second position solution; calculating a first sensor weight of the first position solution based on a likelihood function; calculating a second sensor weight of the second position solution based on the likelihood function; resampling the first position solution based on the first sensor weight to obtain a first resampled position solution; resampling the second position solution based on the second sensor weight to obtain a second resampled position solution; and calculating a final position estimate based on the GPS measurement, the first resampled position solution, and the second resampled position solution.