Virtual Ambient Temperature Sensing for Adaptive Thermal Throttling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current thermal management systems in devices rely on static throttling policies and dedicated sensors, which can lead to inaccurate ambient temperature measurements and unnecessary resource throttling, especially in varying environmental conditions, resulting in degraded user experience and increased costs due to the use of multiple thermistors.
Innovation Solution
The implementation of neural network models, specifically regression and classification neural networks, to estimate the contribution of internal and external factors to a device's temperature, allowing for accurate determination of ambient temperature without dedicated sensors, thereby optimizing thermal management and reducing costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If dedicated ambient temperature sensors are used to measure ambient temperature, then ambient temperature measurement is achieved, but measurement precision deteriorates because the measurement is influenced by hardware component heat contributions
Solution Approach 1:
The patent extracts and isolates the ambient temperature measurement function from the device's internal thermal environment by using machine learning models to separate ambient temperature contribution from hardware-generated heat, allowing accurate ambient temperature sensing without dedicated physical sensors
Solution Approach 2:
The patent creates a virtual ambient temperature sensor using machine learning models that replicate the function of physical ambient temperature sensors by analyzing patterns from multiple existing temperature sensors and usage data to infer ambient temperature
2Adaptability or versatility
If static thermal throttling policies are implemented, then thermal mitigation is achieved, but adaptability deteriorates because the policies do not account for varying ambient temperatures across different regions and seasons
Solution Approach 1:
The patent transforms static thermal throttling policies into dynamic policies that automatically adjust based on inferred ambient temperature, hardware usage patterns, and seasonal variations, allowing the system to adapt thermal management strategies to current environmental conditions
Solution Approach 2:
The patent implements feedback mechanisms where temperature data, usage patterns, and inferred ambient conditions continuously inform and adjust thermal management decisions, creating a closed-loop system that learns and adapts to varying environmental conditions
3Measurement precision
If multiple thermistors are deployed to measure temperatures of different hardware components, then temperature monitoring coverage is improved, but device complexity and cost increase
Solution Approach 1:
The patent makes existing temperature sensors multi-functional by using them both for direct hardware temperature monitoring and as input features for machine learning models that infer ambient temperature, eliminating the need for separate dedicated ambient temperature sensors
Solution Approach 2:
The patent enables the device's existing temperature sensing infrastructure to serve dual purposes: monitoring internal hardware temperatures and inferring external ambient temperature through machine learning analysis, making the system self-sufficient without additional sensors
Data Source
AI summary
Methods and systems for ascertaining factors contributing to the temperature of a device. A method includes monitoring a plurality of parameters that are contributing to a temperature of the device. The method also includes estimating a degree of contribution of internal factors to the temperature of the device based on the monitored plurality of parameters and a battery temperature of a battery of the device. The method further includes estimating a degree of contribution of external factors to the temperature of the device, based on the monitored plurality of parameters and a battery temperature of a battery of the device. A neural network can be used for estimating the temperature of the ambience of the device and the impacts of internal and external factors on temperature of the device.