In-Memory Neural Processing for ADAS Sensor Data Offloading
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in efficiently processing large quantities of sensor data from various sources, leading to processor overload and potential safety hazards due to delayed or missed detection of unsafe driving conditions.
Innovation Solution
A memory device with a processing engine offloads computations from the vehicle's processor, using a Compute Express Link (CXL) device to share only critical decision-making information and store raw data for analysis by a neural network, reducing the processing burden and enabling faster response times for safety actions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the processor processes all sensor data directly, then detection accuracy is improved, but processor load increases leading to delayed response
Solution Approach 1:
The patent segments the processing workload by introducing a memory device with processing capabilities that handles initial sensor data filtering and preprocessing. This divides the overall processing task between the memory device (handling raw data filtering) and the main processor (handling complex analysis), thereby maintaining detection accuracy while reducing processor load and response time.
Solution Approach 2:
The memory device acts as an intermediary between sensors and the main processor. It performs intermediate processing operations including data filtering, validation, and preliminary analysis, which reduces the burden on the main processor and enables faster overall system response while maintaining accurate object detection.
2Reliability
If all sensor data is processed by the main processor, then comprehensive analysis is achieved, but system complexity and processing burden increase
Solution Approach 1:
The processing functions are segmented between the memory device and main processor. The memory device handles data filtering, validation, and preliminary processing, while the main processor focuses on complex object recognition and decision-making. This segmentation maintains comprehensive analysis capability while distributing processing burden.
Solution Approach 2:
The memory device provides self-service processing by autonomously filtering and preprocessing sensor data before it reaches the main processor. This self-service capability reduces the main processor's workload and simplifies the overall system architecture by eliminating the need for the processor to handle every raw data point.
3Productivity
If the processor handles all computations, then processing completeness is maintained, but processing speed decreases due to overload
Solution Approach 1:
The patent extracts computationally intensive preprocessing operations from the main processor and relocates them to the memory device. This extraction of filtering, validation, and preliminary processing functions from the processor enables faster overall processing speed while the processor maintains completeness by focusing on critical decision-making tasks.
Solution Approach 2:
The memory device performs preliminary processing actions on sensor data before it reaches the main processor. By pre-filtering and validating data in advance, the system achieves faster processing speed at the memory device level, while the processor receives already-prepared data for quicker final analysis and decision-making.
Data Source
AI summary
Systems, methods, and apparatus related to memory devices. In one approach, a memory device stores data received from sensors of a vehicle. The memory device includes an offloading engine used to perform computations on portions of the stored data. The offloaded computations include processing using a neural network. An output from the neural network generates an instruction for a processor of the vehicle. The processor of the vehicle reads instruction from a cache of the memory device. Based on the read instruction, the processor of the vehicle controls one or more vehicle functions such as braking, etc.