Radar Tracking Association Using Kinematics Priors to Cut Compute Load
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face high computational loads due to intensive processing of radar signals, which is more accurate than other sensors like cameras or LiDAR, necessitating a method to leverage sensor strengths while minimizing processing load.
Innovation Solution
A method that focuses radar processing on specific locations around the vehicle using data from other sensors like cameras or LiDAR to identify which radar data to process, ignoring signals from non-tracked objects, thereby reducing computational tasks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If radar signals are processed intensively to improve tracking accuracy, then measurement precision is improved, but computational load increases
Solution Approach 1:
The patent segments the tracking process into multiple stages: initial detection by non-radar sensors, candidate target identification, radar signal processing only for candidates, and association refinement. This segmentation allows intensive radar processing to be applied only to a small subset of potential targets rather than all detected objects, thereby maintaining high tracking accuracy while reducing overall computational load.
Solution Approach 2:
The patent applies different processing qualities to different targets based on their likelihood of being actual threats. High-priority candidates identified through sensor fusion and kinematic analysis receive intensive radar processing, while low-priority targets receive minimal or no radar processing. This local differentiation of processing quality maintains accuracy for critical targets while reducing computational burden overall.
2Reliability
If radar data from all objects is processed, then tracking completeness is improved, but processing time increases
Solution Approach 1:
The patent performs preliminary filtering and candidate identification using data from cameras and LiDAR before processing radar signals. By pre-identifying which objects are likely targets based on non-radar sensor data and kinematic priors, the system prepares a shortlist of candidates that require radar processing. This preliminary action ensures that no actual targets are missed (maintaining completeness) while avoiding unnecessary radar processing of irrelevant objects (reducing processing time).
Data Source
AI summary
The present disclosure is directed to focusing processing resources of a radar apparatus to radar data associated with specific locations around an autonomous vehicle (AV). Data associated with a type of sensing apparatus that is not a radar (e.g. a camera and/or a LiDAR apparatus) that tracks motion of specific objects around the AV may be used by a processor to identify specific sets of radar data to process. Locations associated with a specific set of tracked objects may be used to identify sets of received radar signals that will be processed by the radar apparatus. Radar signals received by the radar apparatus that are associated with locations that do not correspond to a location of a tracked object may be ignored by the radar apparatus to reduce a number of computational tasks that a processor of the radar apparatus must process.