Nonlinear State Estimation Using Concurrent Filters and Neural Inference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional inferential sensing methods struggle to accurately estimate the time-varying parameters of complex, highly nonlinear systems, particularly in applications requiring adaptive control, such as autonomous navigation and spacecraft operations, due to underconstrained estimation problems and the lack of generalized parameter estimation methods.
Innovation Solution
A two-step process involving a filter operating in Failure Mode 2 and a multiple concurrent filter methodology, followed by an auxiliary neural network, to estimate time-varying parameters of nonlinear systems, allowing for unlimited parameter estimation independent of the number of system outputs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional inferential sensing methods (e.g., Kalman Filter variants) are used for parameter estimation, then the method is simple and computationally efficient, but it can only handle systems with a limited number of parameters (less than or equal to the number of independent measured outputs) and cannot accurately estimate parameters in highly nonlinear underconstrained systems
Solution Approach 1:
The patent segments the parameter estimation problem by dividing it into two distinct phases: an offline training phase where neural networks are trained using data from conventional filters, and an online execution phase where the trained networks perform rapid parameter estimation. This segmentation allows the system to leverage the simplicity of conventional filters for data generation while achieving the adaptability of neural networks for handling highly nonlinear underconstrained systems.
Solution Approach 2:
The patent performs preliminary action by pre-training neural network models offline using extensive simulation data or experimental data from conventional inferential sensing methods. This preliminary training phase prepares the neural networks to handle complex nonlinear relationships before actual real-time operation, enabling the system to achieve high adaptability without compromising online computational efficiency.
2Measurement precision
If the number of time-varying parameters to be estimated increases beyond the number of independent measured outputs, then the estimation problem becomes underconstrained and more challenging, but conventional methods cannot provide accurate solutions for such cases
Solution Approach 1:
The patent introduces neural networks as intermediary components that bridge the gap between limited measured outputs and the larger number of parameters to be estimated. These neural networks act as mediators that learn complex nonlinear relationships from training data, enabling accurate parameter estimation even when the number of parameters exceeds the number of independent measurements.
Solution Approach 2:
The patent transforms the estimation problem by changing the parameters of the neural network models during offline training. By adjusting the neural network architecture, training data composition, and model parameters during the offline phase, the system optimizes its ability to handle underconstrained problems with higher measurement precision.
3Adaptability or versatility
If specialized ad hoc methods are used for highly nonlinear systems, then the method can provide limited scope adaptive control solutions, but it lacks generality and cannot be broadly applied to different systems
Solution Approach 1:
The patent creates a universal estimation framework that can be applied to various types of highly nonlinear systems across different domains. The neural network-based approach, trained offline with system-specific data, provides a multi-functional solution that maintains reliability for adaptive control while being broadly applicable to different systems, replacing the need for specialized ad hoc methods for each application.
Data Source
AI summary
The embodiments are directed to an inferential sensing system, methods and computer program product of an estimator for estimating parameters of complex nonlinear time-varying systems from scarce system output measurements. The estimator comprises a two-step process to accurately estimate the time-varying parameters of the time-varying system based on the input and output sample of the time-varying system. First, multiple filters in the high frequency processing loop, operating independently and concurrently process the input and output samples of the time-varying system to generate a hypersurface comprising time series objects. Each filter is restricted to adapt only a subset of the modeled time-varying parameters. The hypersurface comprising the time series objects is aggregated over several iterations of the high frequency processing loop. Second, the hypersurface is passed through a neural network in the low frequency processing loop to infer estimates of the time-varying system parameters.