Radar Target Tracking via Linear Gaussian State Equation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current three-dimensional radar target tracking methods face challenges with high computational complexity, limited scalability, and instability, especially in environments with large observation errors, making real-time tracking of moving targets difficult and inaccurate.
Innovation Solution
A method that constructs a state vector and motion model using three-dimensional radar observation data within a linear Gaussian framework, employing dimension-expansion processing and fusion filtering to achieve accurate and robust tracking without increasing computational complexity, suitable for both three-dimensional and two-dimensional scenarios.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep neural networks and machine learning algorithms are used for three-dimensional radar target tracking, then tracking accuracy may be improved, but computational complexity increases significantly
Solution Approach 1:
The patent replaces complex machine learning algorithms with a simplified analytical state equation approach. Instead of using deep neural networks that require extensive computation and training data, the invention uses a mathematical model based on radar observation data that directly calculates target state through dimension-expansion processing and fusion filtering, achieving comparable accuracy with much lower computational complexity
Solution Approach 2:
The patent transforms the tracking problem by changing the parameter representation from traditional three-dimensional Cartesian coordinates to a dimension-expanded state space that includes distance, azimuth, pitch angle and their derivatives. This parameter transformation enables the use of simple linear Gaussian filtering methods instead of complex nonlinear machine learning algorithms
2Productivity
If classical nonlinear filtering methods are used for three-dimensional radar target tracking, then tracking may be achieved, but stability deteriorates in scenes with long distance and large observation error
Solution Approach 1:
The patent replaces unstable classical nonlinear filtering methods with a stable analytical state equation approach. By formulating the tracking problem as a linear Gaussian system through dimension-expansion processing, the invention achieves numerical stability even in challenging scenarios with long distance and large observation errors, eliminating the divergence problems of traditional nonlinear filters
Solution Approach 2:
The patent introduces dimension-expanded state variables as intermediaries between raw radar observations and target state estimation. This intermediate representation transforms nonlinear relationships into linear ones, allowing the use of stable Kalman filtering techniques while maintaining accuracy in difficult tracking scenarios
3Measurement precision
If machine learning methods are used for target tracking, then tracking performance may be improved, but real-time capability is compromised due to high computational complexity
Solution Approach 1:
The patent replaces computationally intensive machine learning methods with efficient analytical calculations. The dimension-expanded state equation approach uses simple matrix operations and linear filtering that can be executed in real-time, eliminating the delay and high computational burden associated with machine learning inference while maintaining tracking performance
Solution Approach 2:
The patent segments the tracking problem into distinct computational components: dimension-expansion processing, state prediction, and fusion filtering. Each component uses efficient mathematical operations that can be computed rapidly, enabling real-time processing whereas machine learning methods treat the entire problem as a monolithic computation
Data Source
AI summary
A method, device, system, and storage medium for tracking a moving target are provided. The method uses three-dimensional radar observation data to construct a state vector and a motion model of the moving target, thereby to construct a state equation and an observation equation for achieving filtering and tracking within a linear Gaussian framework. The disclosure is also suitable for a moving target in a two-dimensional scene with a distance and an azimuth, and the disclosure use a two-dimensional observation vector to construct a dynamic system to achieving tracking of the moving target. The disclosure can be used in radar systems containing Doppler measurements, and tracking of moving targets can be implemented by performing dimension-expansion processing on observation equations.