Terrain Matching Navigation Filter Switching Strategy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current navigation filters used in terrain correlation navigation systems, such as Extended Kalman Filters, are not robust and can lead to filter divergence when faced with weak information, resulting in periods of 'silence' during which the carrier's navigation status is unknown, and particulate filters suffer from degenerative effects over long navigation times.
Innovation Solution
A navigation filter comprising a convergence filter of particulate type and a tracking filter, such as an Extended Kalman Filter or Unscented Kalman Filter, is used to estimate the kinematic state of a carrier, with the selection between these filters based on a quality index calculated from the covariance matrix, allowing for robustness and convergence performance without prolonged 'silence' periods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If an Extended Kalman Filter is used for navigation filtering, then the tracking performance is improved, but the filter divergence occurs when faced with weak information
Solution Approach 1:
The navigation filter dynamically switches between Extended Kalman Filter mode and particle filter mode based on the calculated quality index. When the quality index indicates weak terrain information, the system transitions to particle filter mode to avoid divergence, and switches back to EKF mode when quality improves, making the filter adaptive to varying information quality conditions
Solution Approach 2:
A quality index calculation mechanism serves as an intermediary between the terrain sensor measurements and the navigation filter. This intermediary evaluates the quality of terrain information and controls the switching between different filter types, preventing the EKF from processing low-quality data that would cause divergence
2Reliability
If a particle filter is used for navigation filtering, then the convergence performance is improved, but degenerative effects occur over long navigation times
Solution Approach 1:
The system periodically evaluates the quality index and switches between filter types accordingly. Particle filter mode is activated periodically when quality index indicates weak terrain information, while EKF mode is used during periods of strong terrain information, creating a periodic switching pattern that leverages the strengths of each filter type at appropriate intervals
Solution Approach 2:
The filter type is dynamically adjusted based on the current quality index rather than using a fixed filter type. This dynamic switching allows the system to use particle filters for convergence when needed and EKF for continuous tracking when terrain information is strong, optimizing performance throughout the navigation duration
3Adaptability or versatility
If terrain correlation navigation is used, then autonomous navigation is achieved, but filter divergence leads to periods of silence where navigation status is unknown
Solution Approach 1:
The quality index calculation provides continuous feedback about the state of terrain information quality. This feedback controls the switching between filter types, ensuring that the system maintains navigation status information by switching to particle filter mode when the quality index indicates potential divergence conditions, thereby preventing loss of navigation status
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The navigation filter (11) has a selectively tilting unit for calculating a kinematic state of a carrier and a covariance matrix, where the navigation filter restores values ??calculated by one of a convergence filter (21) e.g. particle filter or marginalised particle filter, and a tracking filter (22) e.g. extended Kalman filter or Unscented Kalman filter. The navigation filter restores the values based on the comparison of a quality index calculated from the covariance matrix restored by the navigation filter at one instant at a predetermined threshold value.