Heterogeneous AI Processor Wake-Up Inference for Low-Power Sensing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing electronic devices suffer from unnecessary power consumption due to unused hardware components being powered on, and there is a lack of precise sensing data recognition leading to incorrect function performance.
Innovation Solution
A trained AI recognition model using an artificial neural network is employed to accurately determine when to perform specific functions by understanding sensing data, reducing power consumption by only activating the system when necessary.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If power is turned on for all hardware components to ensure system availability, then system reliability is improved, but power consumption increases
Solution Approach 1:
The patent applies preliminary action by pre-training an AI recognition model offline to accurately predict when specific functions should be performed. This pre-processing enables the system to make informed decisions about hardware activation before actual operation, ensuring reliability only when necessary and reducing unnecessary power consumption during idle states
Solution Approach 2:
The patent implements dynamics by dynamically adjusting hardware component states based on real-time sensing data and AI model predictions. The system transitions between active and inactive states for different hardware components according to actual needs, rather than maintaining a static power state, thereby optimizing the balance between availability and power consumption
2Measurement precision
If sensing data is continuously monitored to accurately determine function performance timing, then functional precision is improved, but processing time increases
Solution Approach 1:
The patent applies preliminary action by performing extensive AI model training offline before deployment. The pre-trained model contains learned patterns for accurately interpreting sensing data, enabling fast real-time inference without requiring complex processing during actual operation, thus achieving both high precision and low latency
3Measurement precision
If AI model training is performed in real-time to adapt to new sensing data, then model accuracy is improved, but computational overhead increases
Solution Approach 1:
The patent applies preliminary action by separating the training and inference phases. The AI model is trained comprehensively offline using large datasets to achieve high accuracy, then the trained model is deployed for efficient real-time inference. This approach eliminates the need for continuous real-time training, significantly reducing computational overhead and power consumption during device operation while maintaining model accuracy
Data Source
AI summary
A heterogeneous processor includes a first processor and a second processor of a different type. The heterogeneous processor operates in either a low-power mode or a full-power mode. The first processor is configured to operate in the low-power mode, process sensing data from a sensor using a trained neural network model, and generate a wake-up signal when an output of the trained neural network model satisfies a predefined criterion. The wake-up signal is provided to the second processor during the low-power mode. The second processor remains in a powered-down state during the low-power mode and transitions to the full-power mode in response to the wake-up signal.