Hybrid State Estimation Control Under Noisy Sensor Dynamics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining the state of a physical system from sensor data, such as Kalman filtering, face limitations in accuracy due to noise and insufficient modeling of complex dynamics, especially when dealing with noisy measurements from sources like GPS or temperature sensors.
Innovation Solution
A hybrid approach combining a prior knowledge-based mathematical model with a learned model, where the learned model corrects the initial inference from the mathematical model using ground truth data, improving the accuracy of state inference by accounting for unknown relations and correlations not represented in the prior knowledge-based model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Kalman filtering is used to estimate state from sensor measurements, then the state can be determined recursively, but the measurement precision and reliability are degraded due to noise and insufficient modeling of complex dynamics
Solution Approach 1:
The patent combines a learned model (neural network) with a prior knowledge-based mathematical model (such as Kalman filter) to create a hybrid state estimation system. The learned model captures complex dynamics and nonlinear relationships from training data, while the mathematical model provides physical constraints and prior knowledge. Their outputs are merged to produce the final state estimate, achieving both accuracy and robustness.
Solution Approach 2:
The patent creates a composite estimation approach by integrating two different modeling paradigms: data-driven learned models and physics-based mathematical models. This composite approach leverages the strengths of both methods - the learned model's ability to capture complex patterns and the mathematical model's interpretability and physical consistency - to achieve superior state estimation performance.
2Adaptability or versatility
If a prior knowledge-based mathematical model is used to model the state, then physical laws and prior knowledge are incorporated, but the model cannot account for unknown relations and correlations in complex dynamics
Solution Approach 1:
The patent performs preliminary learning during an offline training phase where the learned model is trained on historical sensor data and ground truth states. This preliminary action allows the system to capture complex dynamics and relationships before actual operation, so that during runtime, the hybrid model can accurately infer states without needing to explicitly model all complex relationships.
Solution Approach 2:
The learned model acts as an intermediary that bridges the gap between sensor measurements and true state. It processes the measurements and provides corrections or supplements to the mathematical model's estimates, capturing relationships that the prior knowledge-based model cannot represent, thereby improving overall accuracy.
Data Source
AI summary
A system and computer-implemented method are provided for enabling control of a physical system based on a state of the physical system which is inferred from sensor data. The system and method may iteratively infer the state by, in an iteration, obtaining an initial inference of the state using a mathematical model representing a prior knowledge-based modelling of the state, and by applying a learned model to the initial inference of the state and the sensor measurement, wherein the learned model has been learned to minimize an error between initial inferences provided by the mathematical model and a ground truth and to provide a correction value as output for correcting the initial inference of the state of the mathematical model. Output data may be provided to an output device to enable control of the physical system based on the inferred state.