Radar Motion Subsets for Extended Object Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vehicle environment detection systems face challenges in robustly tracking the motion states of multiple moving and stationary extended objects due to noise sensitivity and error-prone fitting algorithms, particularly in distinguishing and separating targets with similar motion vectors.
Innovation Solution
A vehicle environment detection system employing a radar sensor arrangement and processing unit that generates detection lists including range, azimuth angle, and Doppler velocity, aggregates these into consistently moving motion subsets, and uses non-parametric object modeling to track arbitrary shaped objects without fitting parametric models, thereby enhancing robustness against noise and errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If parametric model fitting algorithms are used to track objects, then motion states can be estimated, but the system becomes sensitive to noise and error-prone
Solution Approach 1:
The patent changes the fundamental parameter representation from parametric models (requiring fitting) to non-parametric point cloud representations. By representing objects as collections of 3D points with associated motion vectors rather than fitting them to predefined geometric models, the system eliminates fitting errors and improves robustness against noise while maintaining motion state estimation accuracy
Solution Approach 2:
The patent extracts and removes the problematic parametric model fitting step from the tracking pipeline. By directly working with raw radar detection points and assigning motion vectors to individual points rather than fitting parametric models, the system eliminates the source of fitting errors and noise sensitivity
2Quantity of substance
If multiple targets with similar motion vectors are tracked, then comprehensive scene monitoring is achieved, but target separation and identification become difficult
Solution Approach 1:
The patent applies segmentation by dividing the point cloud into distinct motion subsets, where each subset corresponds to a separate target object. By clustering points with similar motion characteristics and spatial positions into separate groups, the system achieves effective target separation even when motion vectors are similar, allowing comprehensive monitoring of multiple objects while maintaining identification accuracy
Solution Approach 2:
The patent introduces spatial position as an additional dimension for target separation beyond motion vectors. By combining motion vector similarity with spatial clustering in the segmentation process, the system can distinguish between multiple targets that have similar motion characteristics but different positions, thereby improving target separation accuracy
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system effectively obtains motion states for multiple extended objects, providing robustness against noise and error-prone algorithms, allowing for flexible tracking of objects of any shape and size, and improving object separation and identification.
Implementation Method 1
a radar sensor arrangement (3) and a processing unit (4), where the radar sensor arrangement (3) is arranged to detect at least two radar detections (9, 10, 11) during at least two radar cycles
Implementation Method 2
For each radar cycle, the processing unit (4) is arranged to generate a detection list including range, azimuth angle and Doppler velocity for each one of the radar detections (9, 10, 11)
Data Source
AI summary
A vehicle environment detection system (2) that includes at least one radar sensor arrangement (3) and at least one processing unit (4), where the radar sensor arrangement (3) is arranged to detect at least two radar detections (9, 10, 11) during at least two radar cycles. For each radar cycle, the processing unit (4) generates a detection list ({Dit}t=t<sub2>0</sub2>, . . . ,t<sub2>0</sub2>−N) including range (ri), azimuth angle (φi) and Doppler velocity (vi) for each of the radar detections (9, 10, 11). The processing unit (4) is further arranged to aggregate and store detection lists ({Dit}t=t<sub2>0</sub2>, . . . ,t<sub2>0</sub2>−N) from the radar cycles in a detection memory (12), and then to group the radar detections (9, 10, 11) in the detection lists ({Dit}t=t<sub2>0</sub2>, . . . ,t<sub2>0</sub2>−N) into consistently moving motion subsets (40, 41, 42) in a segmentation procedure. Each motion subset (40, 41, 42) corresponds to a certain target object (6, 7, 8).