Neural Network Power Mode Switching for Battery-Constrained AI Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Portable electronic devices with on-device AI face limitations in power consumption due to battery constraints, necessitating flexible adjustment of resources required to run AI models.
Innovation Solution
An electronic apparatus that uses a neural network model to adaptively adjust power consumption by identifying operation modes based on device state, charging state, and user commands, and adjusts power consumption accordingly by using different neural network models or modifying input features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a neural network model is used to improve device functionality and AI performance, then processing capability and intelligence are improved, but power consumption increases
Solution Approach 1:
The system dynamically adjusts the operation mode of the neural network model based on real-time device states (charging status, temperature, usage patterns). It can switch between different operation modes (e.g., high-performance mode, power-saving mode, temperature-restricted mode) to optimize the balance between AI processing capability and power consumption, making the system adaptable to varying operational conditions rather than operating in a fixed state
Solution Approach 2:
The system changes operational parameters such as input feature selection, model complexity, and processing depth based on device state. When power consumption needs to be reduced, it adjusts parameters like limiting the number of input features, reducing model depth, or lowering processing precision while maintaining acceptable AI performance, thereby controlling power consumption through parameter optimization
2Measurement precision
If more input features are used in the neural network model to improve accuracy, then AI model performance is improved, but power consumption and processing time increase
Solution Approach 1:
The system applies partial action by selectively using only the necessary subset of input features rather than processing all available features. Based on device state and task requirements, it determines the optimal number of input features to use, employing just enough processing power to achieve the required accuracy level without the excessive energy consumption that would result from processing all possible features
Solution Approach 2:
The system applies different processing quality levels to different input features based on their importance and the current operational context. Critical features receive full processing attention while less important features may be processed with reduced precision or excluded entirely, creating a non-uniform processing strategy that optimizes the balance between accuracy and power consumption across different parts of the input data
3Speed
If the device operates in high-performance mode to improve functionality, then processing speed and capability are improved, but battery life decreases
Solution Approach 1:
The system dynamically adjusts operation mode based on battery charge level and usage patterns. When battery charge is high, it can operate in high-performance mode for faster processing. When battery charge drops below thresholds, it automatically transitions to power-saving mode with reduced processing speed, thereby extending battery life while maintaining responsiveness to user needs
Solution Approach 2:
The system periodically monitors device state including battery charge level, temperature, and usage patterns, and adjusts operation mode accordingly. This periodic assessment allows the system to alternate between high-performance and power-saving modes based on current conditions, optimizing the balance between processing speed and battery life over time rather than maintaining a fixed mode
Data Source
AI summary
An electronic apparatus includes memory storing instructions, and at least one processor connected to the memory and configured to control the electronic apparatus using a neural network model, where the instructions, when executed by the at least one processor, cause the electronic apparatus to identify whether to use the neural network model based on a state of the electronic apparatus, based on identifying that the neural network model is to be used, identify an operation mode of the electronic apparatus as one of a first mode having a first power consumption or a second mode having a second power consumption greater than the first power consumption, and adjust power consumption of the electronic apparatus based on the identified operation mode.