Radar ECU Neuromorphic Processing for ADAS Workload Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous driving systems face challenges in efficiently processing and analyzing radar data, leading to increased computational workload and data traffic, which can impact the safety and performance of autonomous vehicles.
Innovation Solution
Implementing a neuromorphic memory device with a Spiking Neural Network (SNN) in a radar Electronic Control Unit (ECU) to analyze radar images, generating inference results that are sent to the Advanced Driver Assistance System (ADAS), thereby reducing data transfer and enhancing processing capabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If radar data is processed centrally by the ADAS processor, then computational accuracy is maintained, but computational workload and data traffic increase
Solution Approach 1:
The patent divides the computational workload by implementing a distributed processing architecture where the radar ECU performs local preprocessing of radar data using a neural network, extracting relevant features and generating inference results. Only these processed results are transmitted to the central ADAS processor, rather than transmitting all raw radar data. This segmentation reduces the computational burden on the central processor while maintaining accurate object detection and classification capabilities.
2Loss of energy
If radar data is processed locally at the ECU, then data transfer is reduced, but processing capability must be enhanced
Solution Approach 1:
The patent introduces a neural network as an intermediary processing layer within the radar ECU. This neural network acts as a mediator that performs intelligent preprocessing of radar data, identifying objects and generating inference results locally. By placing this intermediary processing capability at the edge (radar ECU), the system reduces the volume of data that needs to be transmitted to the central processor, thereby reducing energy consumption for data transfer while maintaining enhanced processing capabilities through the neural network.
3Measurement precision
If more sensors are installed to improve detection accuracy, then object identification improves, but system complexity and cost increase
Solution Approach 1:
The patent implements a virtual model of object characteristics by using a neural network to generate inference results that replicate the detection and classification capabilities typically requiring multiple physical sensors. The neural network processes radar data to create detailed object models including type, distance, speed, and other attributes, effectively copying the information-gathering function that would otherwise require additional sensors. This approach achieves high measurement precision for object identification while avoiding the complexity and cost of installing and integrating multiple sensor types.
Data Source
AI summary
Systems, methods and apparatuses of radar Electronic Control Units (ECUs) of autonomous vehicles. A radar ECU can include: a memory configured to store a radar image and an Artificial Neural Network (ANN); an inference engine configured to use the (ANN) to analyze the radar image and generate inference results; and a communication interface coupled to a computer system of a vehicle to implement an advanced driver assistance system to operate the controls according to the inference results and a sensor data stream generated by sensors configured on the vehicle.