Automotive Radar Signal Sampling for Bandwidth-Limited Perception
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately classifying and tracking objects using radar signals due to computational, storage, and bandwidth constraints, with existing methods either requiring excessive resources or filtering out valuable information.
Innovation Solution
Selecting a patch of frequency bins based on proximity to peaks in radar signals, applying downsampling, clipping, and compression techniques to transmit only relevant data to a processing unit for machine learning models, reducing computational, storage, and bandwidth requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the entirety of RDA signals is transmitted to the post processing unit, then accurate object detection and classification is achieved, but bandwidth requirements and computational resources exceed typical system constraints
Solution Approach 1:
The patent extracts and transmits only the most relevant portions of the RDA signal data to the post-processing unit. Specifically, it identifies and transmits range-Doppler map data corresponding to detected targets and their immediate surroundings, rather than transmitting the entire raw signal dataset. This extraction approach maintains object detection accuracy while significantly reducing bandwidth requirements.
Solution Approach 2:
The patent segments the RDA signal data into meaningful portions based on detected targets. It divides the range-Doppler map into regions of interest centered around detected objects, transmitting only these segmented portions along with their metadata. This segmentation allows the system to focus computational resources on relevant data while reducing overall data transmission volume.
2Measurement precision
If high-resolution time-frequency analysis is applied to extract Doppler signatures, then accurate Doppler information is obtained, but computational resources and memory requirements are excessive for typical sensors
Solution Approach 1:
The patent performs preliminary Doppler analysis and target detection at the sensor level before transmission. By pre-processing the RDA signals to identify targets and extract their Doppler characteristics, the system prepares the data in advance so that the post-processing unit receives already-processed, target-specific information rather than raw signals requiring intensive computation.
Solution Approach 2:
The patent extracts only the Doppler signature information relevant to detected targets, rather than performing full high-resolution time-frequency analysis on the entire signal dataset. It isolates and transmits Doppler characteristics specifically associated with identified objects, reducing computational requirements while maintaining accuracy.
3Reliability
If CFAR algorithm is applied to RDA signals, then false alarm rate is controlled, but much useful information is filtered out necessary for accurate object detection and classification
Solution Approach 1:
The patent applies CFAR processing selectively to specific regions of the range-Doppler map rather than uniformly across all data. It concentrates false alarm suppression efforts around detected targets where it is most needed, while preserving signal information in other regions that may contain additional targets or important contextual data. This localized approach maintains reliability without excessive information loss.
4Loss of information
If the entire RDA FFT map is transmitted, then full information about the road scene is provided, but bandwidth requirements exceed typical system support
Solution Approach 1:
The patent extracts and transmits only the essential road scene information contained in the RDA FFT map, specifically focusing on regions containing detected targets and their immediate environments. It removes redundant or less critical data portions, transmitting a condensed representation that preserves key scene information while reducing bandwidth requirements to levels supported by typical automotive systems.
Data Source
AI summary
In various examples, methods and systems are provided for sampling and transmitting the most useful information from a radar signal representing a scene while staying within the computational and storage confines of a standard automotive radar sensor and the bandwidth constraints of a standard communication link between a radar sensor and processing unit. Disclosed approaches may select a patch of frequency bins that correspond to radar signals based at least on proximities of the frequency bins to one or more frequency bins corresponding to at least one peak and/or detection point in the radar signals. Data representing samples corresponding to the patch of frequency bins may be transmitted to the processing unit and applied to one or more machine learning models in order to accurately classify, identify, and/or track objects.