Extended Object Tracking With Center-Truncated Radar Measurements
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automotive object tracking systems using automotive radar measurements struggle to accurately capture the expanded state of objects, including both kinematic and extended states, due to complex real-world measurement distributions and noise, leading to inaccurate tracking and high computational complexity.
Innovation Solution
A system and method employing a probabilistic filter with a center-truncated distribution and underlying Gaussian distribution pair to estimate the expanded state, which includes a kinematic state and extended state, using a hierarchical measurement model that captures spatial distribution and sensor noise, and a Bayesian EOT algorithm for recursive prediction and update of the expanded state and truncation interval.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If contour model or surface model is used to capture spatial distribution of radar measurements, then measurement precision is improved, but device complexity increases due to cumbersome data association requirements
Solution Approach 1:
The patent extracts and removes the cumbersome data association step from the tracking framework by using a point object tracking approach with Gaussian distribution, thereby simplifying the system while maintaining measurement precision through probabilistic filtering
Solution Approach 2:
The patent changes the measurement model from complex spatial distribution models (contour/surface) to a simpler Gaussian distribution model that captures the essential measurement characteristics without requiring complex data association, thus reducing device complexity while maintaining adequate measurement precision
2Measurement precision
If contour model with random hyper surface or Gaussian process is used to model complex shapes, then measurement precision is improved, but device complexity increases due to high computational requirements
Solution Approach 1:
The patent replaces expensive, computationally intensive contour models with a cheaper, simpler Gaussian distribution model that provides adequate tracking performance without requiring high computational resources, effectively using a simpler approximation instead of a complex exact model
3Device complexity
If surface model such as Gaussian-based ellipse is used for tracking, then device complexity is reduced, but measurement precision deteriorates because the model does not accurately represent real-world radar measurements distributed around object edges
Solution Approach 1:
The patent modifies the measurement model by using a Gaussian distribution centered at the object position with covariance representing object extent, changing the parameters from fixed geometric shapes to probabilistic parameters that better capture the distributed nature of radar measurements around object edges
4Measurement precision
If extended object tracking with multiple measurements per scan is implemented, then measurement precision is improved by capturing spatial distribution, but device complexity increases due to non-scalable data association
Solution Approach 1:
The patent removes the data association step entirely by adopting a point object tracking approach where multiple radar measurements are processed as independent observations updated through probabilistic filtering, eliminating the need for complex many-to-many association algorithms
Solution Approach 2:
The patent creates a universal tracking framework that handles both point objects and extended objects using the same probabilistic filtering mechanism, making the system scalable and adaptable to different object types without requiring separate data association algorithms
Data Source
AI summary
A system and a method for tracking an expanded state of an object including a kinematic state indicative of a position of the object and an extended state indicative of one or combination of a dimension and an orientation of the object is provided herein. The system comprises at least one sensor configured to probe a scene including a moving object with one or multiple signal transmissions to produce one or multiple measurements of the object per the transmission, and a processor configured to execute a probabilistic filter tracking a joint probability of the expanded state of the object estimated by a motion model of the object and a measurement model of the object, wherein the measurement model includes a center-truncated distribution having truncation intervals. The system further comprises an output interface configured to output the expanded state of the object.