AI Model Switching in Wireless Terminals for Low-Overhead Channel Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing radio communication technologies lack sufficient studies on life cycle management (LCM) of AI models, leading to inadequate overhead reduction and channel estimation, which hampers communication throughput and quality improvement.
Innovation Solution
A terminal and base station implementation that includes a transmitting section for reporting a supported model identifier, a receiving section for indicating activation, deactivation, or switch of models, and a control section for performing these actions based on the indication, facilitating efficient LCM.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If AI models are utilized for control and management in radio communication systems, then communication throughput and quality can be improved, but overhead and resource consumption increase due to insufficient life cycle management studies
Solution Approach 1:
The patent implements dynamic model management where AI models can be activated, deactivated, switched, or fallen back based on operational conditions. The base station receives indications from the network management entity to dynamically adjust which AI models are active, allowing the system to adapt resource allocation and model usage to current network conditions, thereby optimizing throughput while controlling overhead.
Solution Approach 2:
The patent changes the operational parameters of AI models by introducing model identifiers that indicate different states (activation, deactivation, fallback, switch). By managing models through these parameter changes rather than permanently deploying all models, the system can optimize performance while reducing the overhead associated with maintaining unnecessary AI model instances.
2Reliability
If multiple AI models are deployed for beam prediction and channel estimation, then communication quality improves, but device complexity and resource management burden increase
Solution Approach 1:
The patent introduces a network management entity as an intermediary that centralizes the management of AI models. This entity sends indications to base stations regarding which models should be activated, deactivated, switched, or fallen back. By offloading the management complexity to this intermediary, individual base stations and terminals experience reduced complexity while still benefiting from multiple AI models improving communication quality.
Solution Approach 2:
The patent segments the AI model management function into distinct operational states represented by model identifiers (activation, deactivation, fallback, switch). This segmentation allows the system to manage multiple AI models through discrete, manageable states rather than as a monolithic complex system, reducing the perceived complexity while maintaining the ability to use multiple models for improved reliability.
3Measurement precision
If AI models are continuously updated and managed, then channel estimation accuracy improves, but loss of time and signaling overhead increase
Solution Approach 1:
The patent implements periodic model management where the network management entity sends indications to base stations at appropriate intervals to activate, deactivate, switch, or fallback AI models. This periodic action allows channel estimation accuracy to be maintained through updated models without requiring continuous management overhead, as models are updated at optimized intervals rather than continuously.
Solution Approach 2:
The patent employs preliminary action by having the network management entity pre-determine which AI models should be active based on anticipated network conditions. By sending activation indications in advance rather than reacting to performance degradation, the system maintains accurate channel estimation while minimizing the time and signaling overhead associated with reactive model management.
Data Source
Figure 1
Figure 2A~2D
Figure 3
AI summary
A terminal according to one aspect of the present disclosure includes a transmitting section that performs reporting of a supported model identifier (ID), a receiving section that receives an indication of at least one of activation, deactivation, fallback, and switch of a model, and a control section that performs, based on the indication, at least one of the activation, the deactivation, the fallback, and the switch. According to one aspect of the present disclosure, preferable overhead reduction/channel estimation/resource use can be achieved.