Autonomous Vehicle Radar Computing Architecture for Low-Bandwidth Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional autonomous vehicle (AV) computing architectures face challenges in data transmission and processing efficiency due to the loss of information when transforming radar sensor data into three-dimensional point clouds, resulting in low resolution and insufficient bandwidth for communication.
Innovation Solution
A computing architecture where a local data processing device within the radar system executes a first deep neural network (DNN) to transform and extract embedded features from radar data tensors, reducing the need for transmitting raw data to a centralized processing device, which then executes a second DNN for object detection and labeling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If radar sensor data is transformed into three-dimensional point clouds for processing, then object detection can be performed using deep neural networks, but information is lost and resolution is substantially reduced
Solution Approach 1:
The patent applies preliminary action by performing feature extraction and transformation operations on radar data tensors before transmission to the centralized processing device. The local computing device executes a first deep neural network to extract embedded features and transform the data into a format suitable for the second DNN, thereby preserving critical information that would otherwise be lost in conventional point cloud transformations.
Solution Approach 2:
The patent introduces an intermediary processing stage between raw radar data and the central DNN processing. The local computing device acts as an intermediary that transforms raw radar data tensors into extracted feature representations, which are then transmitted to the centralized device. This intermediary step preserves information while reducing data transmission requirements.
2Loss of information
If raw radar data tensors are transmitted to the centralized processing device, then complete information is available for processing, but communication bandwidth is insufficient
Solution Approach 1:
The patent applies the extraction principle by removing unnecessary data elements from the transmission stream. Instead of transmitting complete raw radar data tensors, the local computing device extracts only the essential embedded features and transforms the data into a compact representation that retains the information needed for accurate object detection while significantly reducing transmission volume.
Solution Approach 2:
The patent changes the parameters of the data being transmitted by transforming it from raw radar tensor format to extracted feature format. This parameter transformation reduces the data volume while preserving the critical information needed for object detection, effectively solving the bandwidth limitation problem.
3Device complexity
If a conventional centralized processing architecture is used, then all processing can be performed by one device, but network load and latency increase
Solution Approach 1:
The patent applies segmentation by dividing the processing architecture into two distinct segments: a local computing device that performs feature extraction and data transformation, and a centralized processing device that performs object detection and scene classification. This segmentation enables parallel processing operations, reducing network load and processing latency while maintaining architectural simplicity.
Data Source
AI summary
The technologies described herein relate to a computing architecture in an autonomous vehicle (AV). The AV includes a radar system that includes a local data processing device. The AV also includes a centralized data processing device. The local data processing device is configured to perform processing on raw sensor data generated by the radar system to form a feature vector. The local data processing device transmits the feature vector to the centralized data processing device, which performs further processing to identify an object in an environment of the AV.