Terminal Capability Reporting for Dynamic Model Resource Adaptation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for reporting terminal device capabilities are based on fixed reference indicators, which fail to account for dynamically changing resource allocation during task execution, leading to inconsistent model performance and potential failure.
Innovation Solution
A method for dynamically reporting first capability information associated with a model, including resource usage and running process information, allowing network devices to adjust models and strategies accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If capability information is reported based on fixed reference indicators at one time, then the reporting complexity is reduced, but the model cannot adapt to dynamically changing resources during actual running
Solution Approach 1:
The capability information is transformed from a static fixed reference indicator to a dynamic reporting mechanism that updates based on actual resource availability. The terminal device now reports capability information that reflects current resource states (computing power, memory, energy) rather than predetermined fixed values, enabling the model to adapt to changing conditions during runtime.
Solution Approach 2:
A feedback loop is established where the terminal device monitors its actual resource usage and capability status during model execution, then reports updated capability information to the network device. This feedback mechanism allows the system to adjust model deployment decisions based on real-time resource availability, resolving the contradiction between simple reporting and adaptive capability.
2Reliability
If capability information is reported dynamically with resource and running process information, then model performance stability is improved, but the communication overhead increases
Solution Approach 1:
The capability information reporting is extracted from a single static value and divided into multiple targeted information elements (resource availability, running process status, capability level). Only the necessary and relevant information components are transmitted, avoiding redundant data while ensuring model performance stability through comprehensive yet concise capability description.
3Ease of operation
If fixed reference indicator capability information is used, then the initial model selection is simplified, but the model fails to maintain normal operation under varying resource conditions
Solution Approach 1:
The system performs preliminary capability assessment using fixed reference indicators for initial model selection, then continuously updates capability information during runtime. This preliminary action followed by dynamic adjustment allows simple initial decision-making while ensuring ongoing model operation reliability through real-time capability monitoring and reporting.
Data Source
AI summary
A method for communication includes: transmitting, by a terminal device, first capability information; where the first capability information is associated with a first model, and the first capability information is used for indicating one piece of the following information: information of a resource capable of being used by the first model, and information of a running process of the first model.