Compound Measurement Model for Automotive Radar Object Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automotive object tracking systems using automotive radar measurements struggle to accurately capture real-world kinematic and extended states of objects due to complex measurement distributions and noise, limiting their ability to provide precise tracking of object dimensions and orientations.
Innovation Solution
A compound measurement model is introduced, which combines probabilistic distributions constrained to the object's contour with a predetermined geometrical mapping, allowing for more flexible and accurate representation of real-world measurements, and is learned offline for online tracking using probabilistic multi-hypothesis tracking methods like unscented Kalman filter-probabilistic multi-hypothesis tracking.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional point object tracking with single measurement per scan is used, then the tracking system is simple, but it only provides kinematic state and cannot capture extended state (dimension and orientation)
Solution Approach 1:
The patent segments the object representation into fixed set of points on a rigid body, where each point can be independently associated with radar measurements. This segmentation allows the system to track extended state (dimension and orientation) by distributing multiple measurements across multiple points, rather than treating the object as a single point.
Solution Approach 2:
The patent transitions from point object tracking (0-dimensional) to extended object tracking by augmenting the object state to include spatial dimensions (length, width, orientation). This dimensionality change enables the system to capture both kinematic state and extended state, providing comprehensive object characterization.
2Measurement precision
If contour model is used to capture measurement distribution along object contour, then extended state tracking is improved, but data association between fixed points and radar detections becomes cumbersome and non-scalable
Solution Approach 1:
The patent applies local quality by assigning different properties to different parts of the object representation. Each fixed point on the rigid body has localized measurement associations, allowing the system to handle complex measurement distributions locally at each point while maintaining overall system scalability. This localizes the data association complexity rather than requiring global complex associations.
3Productivity
If surface model with inner surface of two-dimensional shape is assumed, then computational complexity is reduced, but real-world automotive radar measurements with noise and surface-only reflections are not accurately captured
Solution Approach 1:
The patent changes the parameters of the measurement model by introducing a probabilistic measurement model that accounts for noise and surface-only reflections. Instead of assuming a idealized surface model, the system adjusts the measurement parameters to reflect real-world radar measurement characteristics, including noise distributions and reflection patterns from object surfaces.
Data Source
AI summary
A tracking system for tracking an expanded state of an object is provided. The tracking system comprises at least one processor and a memory having instructions stored thereon that, when executed by the at least one processor, cause the tracking system to execute a probabilistic filter that iteratively tracks a belief on the expanded state of the object, wherein the belief is predicted using a motion model of the object and is further updated using a compound measurement model of the object. The compound measurement model includes multiple probabilistic distributions constrained to lie on a contour of the object with a predetermined relative geometrical mapping to the center of the object. Further, the tracking system tracks the expanded state of the object based on the updated belief on the expanded state.