On-Wafer Sensing and ML Processing for Low-Transmission Edge AI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional edge devices lack integrated sensing and processing capabilities for locally extracting information and features from analog sensing data and performing machine learning processing, leading to the need for digitalization and transmission of large data volumes, which is energy-intensive, time-consuming, and raises privacy concerns.
Innovation Solution
A semiconductor device with integrated sensing and machine learning processing capabilities, featuring a sensing module and a machine learning processor fabricated on a single wafer, including crossbar arrays for preprocessing analog sensing data and an analog-to-digital converter for digital conversion, enabling local data processing and reducing the amount of data transmitted.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sensing data is digitalized and transmitted to a remote computing device for machine learning processing, then machine learning processing capabilities are achieved, but energy consumption and data transmission time increase significantly
Solution Approach 1:
The patent merges the sensing module and machine learning processing unit into a single integrated device fabricated on one wafer. The sensing module generates analog sensing data that is directly processed by the machine learning processing unit through crossbar arrays and analog-to-digital converters, eliminating the need to transmit large volumes of raw sensing data to remote devices. This integration reduces energy consumption associated with data transmission while maintaining machine learning processing capabilities.
2Reliability
If sensing data is transmitted to a remote device for processing, then machine learning analysis is performed, but data transmission time and processing delay increase
Solution Approach 1:
The patent combines the sensing module and machine learning processing unit into a single integrated device. The machine learning processing unit is directly connected to the sensing module, allowing immediate processing of analog sensing data without external transmission. This integration enables real-time or near-real-time machine learning analysis, significantly reducing data transmission time and processing delays.
3Reliability
If raw sensing data is transmitted for processing, then comprehensive machine learning analysis is possible, but privacy concerns increase
Solution Approach 1:
The patent integrates the sensing module and machine learning processing unit into a single device, enabling local processing of sensing data. The machine learning processing unit performs analysis on the sensing data within the same device where the data is generated, minimizing the need to transmit sensitive raw sensing data externally. This local processing approach maintains machine learning analysis capability while reducing privacy risks associated with data transmission and storage.
4Device complexity
If conventional edge devices are used without integrated sensing and processing capabilities, then device simplicity is maintained, but computational capabilities for local ML processing are insufficient
Solution Approach 1:
The patent integrates the sensing module and machine learning processing unit into a single device fabricated on one wafer. The machine learning processing unit includes crossbar arrays for analog matrix multiplication and analog-to-digital converters for data conversion. This integration provides sufficient computational capability for local machine learning processing while maintaining a compact and unified device structure, avoiding the need for separate sensing and processing components.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables efficient, localized machine learning processing of analog sensing data, reducing data transmission and energy consumption while addressing privacy concerns, and facilitating real-time applications.
Implementation Method 1
a sensing module configured to generate a plurality of analog sensing signals
Implementation Method 2
one or more crossbar arrays configured to process the analog sensing signals to generate analog preprocessed sensing data
Implementation Method 3
an analog-to-digital converter (ADC) configured to convert the analog preprocessed sensing data into digital preprocessed sensing data
Data Source
AI summary
The present disclosure provides for a semiconductor device with integrated sensing and processing functionalities. The semiconductor device includes a sensing module configured to generate a plurality of analog sensing signals; and a machine learning (ML) processor. The sensing module and the ML processor are fabricated on a single wafer. The ML processor includes crossbar arrays that processes the analog sensing signals to generate analog preprocessed sensing data; an analog-to-digital converter (ADC) to convert the analog preprocessed sensing data into digital preprocessed sensing data; and a machine learning processing unit to process the digital preprocessed sensing data utilizing one or more machine learning model.