Laminated AI Sensor Processing Across Multiple Sensor Nodes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Laminated AI sensors face issues with image quality degradation due to electromagnetic noise, warpage from thermal expansion, increased thermal noise, and reduced processing speed and capability due to low-spec arithmetic units, limiting their functionality to basic tasks like counting or face recognition.
Innovation Solution
A sensor apparatus with a laminated structure of a first semiconductor layer containing a pixel array unit and a second semiconductor layer with an AI processing unit, where processing is distributed among multiple sensor apparatuses using a control unit, enhancing robustness and processing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If AI processing is performed on the second semiconductor layer in a laminated structure, then real-time processing and privacy protection are achieved, but electromagnetic noise, thermal expansion warpage, and thermal noise increase
Solution Approach 1:
The patent divides the AI processing function into multiple independent processing units distributed across different semiconductor layers. Each unit handles specific processing tasks, allowing the system to achieve real-time processing capability while distributing thermal and electromagnetic interference across separate physical locations, thereby reducing the harmful effects concentrated in a single layer.
Solution Approach 2:
The patent introduces an intermediary heat dissipation structure and signal transmission mechanism between the first and second semiconductor layers. This intermediary layer acts as a buffer that isolates thermal expansion and electromagnetic noise while maintaining the functional connection for AI processing, thus resolving the contradiction between processing speed and harmful factors.
2Use of energy by moving object
If low-spec arithmetic units are used in the AI processing unit, then manufacturing cost and power consumption are reduced, but processing capability and speed are limited
Solution Approach 1:
The patent segments the AI processing workload into multiple independent tasks distributed across several low-spec arithmetic units in the second semiconductor layer. By parallelizing processing across multiple units, the system achieves high overall processing capability and speed while each individual unit consumes minimal power and can be manufactured with lower specifications.
Solution Approach 2:
The patent combines the processing power of multiple low-spec arithmetic units through a coordinated control mechanism, achieving the equivalent computational capability of a single high-spec unit. This merging approach allows the system to maintain high AI processing capability while using energy-efficient, low-power components that reduce manufacturing costs and power consumption.
3Reliability
If processing is distributed among multiple sensor apparatuses, then robustness and processing capability are improved, but system complexity and communication overhead increase
Solution Approach 1:
The patent segments the AI processing system into multiple independent sensor apparatuses, each capable of autonomous processing. This segmentation improves robustness by distributing processing load and providing redundancy, while the modular architecture actually simplifies system management compared to a centralized approach, as each unit operates independently with standardized interfaces.
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
The distributed processing improves AI processing robustness and speed, allowing for more complex image analysis tasks beyond basic recognition, while reducing the risk of data leakage and power consumption.
Implementation Method 1
a pixel array unit in which a plurality of pixels each having a light receiving element are two-dimensionally arranged
Data Source
AI summary
For a sensor apparatus as a laminated AI sensor, the processing efficiency of an AI processing unit is improved, so that the AI processing function is enhanced, the speed is increased, and the robustness of the AI processing is improved.A sensor apparatus according to the present technology includes a structure in which a first semiconductor layer including an image array unit in which a plurality of pixels having a light receiving element is two-dimensionally arranged, and a second semiconductor layer including an AI processing unit that performs AI processing, which is processing using an AI model, on a captured image obtained on the basis of a light receiving operation by the pixel array unit are laminated, and a control unit that performs control for distributing processing executed by the AI processing unit with another sensor apparatus having a structure in which the first semiconductor layer and the second semiconductor layer are laminated.