Parallel INS Filters for Solution-Domain Instability Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Inertial navigation systems (INS) face instability due to undetected errors in inertial measurement unit (IMU) measurements, leading to inaccurate GNSS/INS solutions, as conventional methods fail to identify unstable filters effectively.
Innovation Solution
Configuring multiple INS filters differently to compute misclosure values, generate rolling histories, and calculate residual and ratio values to determine filter stability, allowing for switchover to a stable filter when instability is detected.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional single-filter approaches are used, then device complexity is reduced, but filter stability and solution accuracy deteriorate due to undetected errors
Solution Approach 1:
The system divides the single filter into multiple parallel filters (e.g., first and second INS filters) with different configurations. Each filter processes the same IMU and GNSS data independently, allowing comparison of their outputs to detect instability. This segmentation enables error detection through diversity while maintaining manageable complexity through modular architecture.
Solution Approach 2:
The multiple filters are configured with different parameters (e.g., different process noise models, different update weights, or different state vectors) to produce varied solutions. By changing configuration parameters across filters, the system creates detectable differences when one filter becomes unstable, enabling reliability monitoring without requiring fundamentally different filter structures.
2Reliability
If multiple filters with different configurations are deployed, then filter stability detection improves, but device complexity increases
Solution Approach 1:
The multiple filters serve dual purposes: they simultaneously generate navigation solutions and perform mutual stability validation. The same filter infrastructure that produces the GNSS/INS solution also generates the reference solutions needed for comparison, eliminating the need for separate validation hardware and reducing overall system complexity.
Solution Approach 2:
The system implements feedback by continuously comparing solutions from multiple filters and using the comparison results to detect instability. The misclosure values and residual analyses create a feedback loop that monitors filter health in real-time, allowing the system to identify and switch away from unstable filters while maintaining continuous navigation functionality.
3Measurement precision
If error detection is performed in the observation domain, then measurement accuracy is maintained, but filter instability remains undetected leading to solution inaccuracies
Solution Approach 1:
The system adds a new dimension of analysis by performing error detection in the solution domain rather than solely in the observation domain. By comparing the actual filter solutions against expected solutions from parallel filters, the system creates a new detection dimension that identifies instability that conventional observation-domain methods miss, while maintaining measurement precision through multi-dimensional validation.
Data Source
AI summary
Techniques are provided for analyzing multiple filters in a solution domain to identify filter instability. A processor may compute, for each filter, a residual value for each of one or more update types utilizing misclosure values that correspond to the update type and are from a rolling history. The processor may compute a ratio value from the residual values computed for the same update type and across two different filters. The ratio value may indicate which filter is experiencing a larger magnitude of difference between two solutions utilized to generate a combined solution. The processor may compare the ratio value, over a time period, with one or more threshold values to identify at least one unstable filter. In response, the processor may take one or more actions, e.g., using the solution from a stable filter instead of the unstable filter, re-initializing the unstable filter using the stable filter, etc.