Calibration of a multi-sensor fusion system for inertial navigation
The method optimizes covariance matrices for extended Kalman filters using natural gradient descent to enhance the precision and accuracy of multi-sensor fusion systems, addressing the limitations of existing technologies in modeling sensor behaviors and environmental adaptability.
Patent Information
- Application Number
- FR2023004486
- Authority / Receiving Office
- FR · FR
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-05-04
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2043-05-04
AI Technical Summary
Existing multi-sensor fusion systems for inertial navigation, such as those using Kalman filters, face challenges in accurately modeling sensor behaviors due to restrictive linear assumptions and convergence issues in optimizing covariance matrices, leading to suboptimal precision and accuracy in navigation estimation.
The method employs an extended Kalman filter calibrated using natural gradient descent to optimize covariance matrices, either statically or dynamically, ensuring accurate modeling of uncertainties and adapting to environmental changes through factorization and neural network-based adjustments.
This approach enhances the precision and accuracy of navigation estimation by effectively quantifying and adapting uncertainty modeling, improving the performance of multi-sensor fusion systems in various environments.
Smart Images

Figure 00000000_0000_ABST
Abstract
Citation Information
Patent Citations
Non-discoloring [hokishimu[hokishimu] emulsion and manufacturing method thereof
JP1989000010A
