Radar Object Tracking Using Innovation Vector Thresholds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Advanced driver-assistance systems (ADAS) face challenges in accurately tracking objects using radar data, particularly due to false associations and erroneous updates, which can lead to incorrect target selection for safety features like automatic emergency braking (AEB).
Innovation Solution
A computer-implemented method using a Kalman-filter method to track objects with radar measurement data, where an innovation vector is computed and compared against predefined thresholds to mark objects as reliable or unreliable, thereby detecting and correcting erroneous updates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If location radar sensors are used to provide data from the surroundings, then the coverage and detection capability are improved, but the number of point-like reflections increases leading to false associations
Solution Approach 1:
The patent implements a feedback mechanism by continuously monitoring the innovation vector magnitude and comparing it against dynamically adjusted thresholds. When the innovation vector exceeds the upper threshold, the system marks the association as unreliable and adjusts thresholds accordingly, creating a closed-loop system that adapts to varying measurement conditions and reduces false associations
Solution Approach 2:
The patent changes the parameter of threshold values dynamically based on the innovation vector magnitude. By adjusting the first and second thresholds according to the observed innovation values, the system adapts its association criteria to current measurement conditions, thereby improving reliability without sacrificing detection capability
2Loss of information
If multiple sensors are used to provide data from the surroundings, then the information completeness is improved, but the association problem becomes more difficult leading to false associations
Solution Approach 1:
The patent introduces the innovation vector as an intermediary metric that simplifies the association problem. Instead of directly comparing multiple sensor measurements with tracked objects, the system uses the innovation vector magnitude as an intermediate indicator of association quality, making the multi-sensor association process more manageable and less prone to errors
3Duration of action of stationary object
If false associations occur in tracking, then the tracking continuity is maintained, but the object state is steered away from the true state leading to incorrect target selection
Solution Approach 1:
The system uses feedback from the innovation vector magnitude to detect when associations are likely false. By continuously monitoring this metric and comparing against thresholds, the system can identify when tracking continuity is being maintained at the cost of accuracy, and take corrective action by marking associations as unreliable
Solution Approach 2:
The patent applies preliminary anti-action by proactively marking associations as unreliable before they can significantly steer the object state away from the true state. By using the innovation vector threshold comparison in advance, the system prevents false associations from corrupting the tracking data, rather than correcting them after damage is done
Data Source
Figure 1
Figure 2
AI summary
A computer-implemented method for tracking objects comprises tracking at least one object with a Kalman-filter method, using radar measurement data, wherein the Kalman-filter method comprises computing an innovation vector based on the radar measurement data. For each tracked object, a component of the innovation vector corresponding to the tracked object is stored. For each tracked object, an updated value of the innovation vector is compared with a predefined first threshold. If the updated value of the innovation vector is above the first threshold, the tracked object is marked as unreliable and the first threshold is stored as a new value of the component of the innovation vector. For each tracked object, an updated value of the innovation vector is compared with a predefined second threshold smaller than the first threshold. If the updated value of the innovation vector is below the second threshold, an unreliable tracked object is marked as reliable and the second threshold is stored as a new value of the component of the innovation vector.