UE AI Capability Reporting for Random Access Resource Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge lies in how a base station learns relevant parameters of an AI network model at a User Equipment (UE), particularly in optimizing resource allocation based on UE's AI network model capabilities.
Innovation Solution
A method for a UE to report its AI network model support capability using uplink resources during a random access procedure, allowing the base station to configure optimal demodulation reference signals based on the UE's capabilities, thereby saving resources and signaling overhead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the terminal determines support capability based on multiple AI framework types and network model types, then the support capability information becomes comprehensive and accurate, but the capability report length increases and may exceed the maximum length of the capability indicator field
Solution Approach 1:
The capability report is divided into two parts: a capability indicator field for common framework types shared by multiple network model types, and a capability indicator extension field for specific framework types of specific network model types. This segmentation allows comprehensive reporting while managing field length constraints.
Solution Approach 2:
The patent adds a new dimension (extension field) to the capability report structure, transitioning from a single flat field to a two-level hierarchical structure with base indicators and extension indicators, enabling expanded information capacity without increasing the base field length.
2Adaptability or versatility
If the terminal reports all supportable AI frameworks and network models, then the network device can fully utilize terminal capabilities, but the reporting overhead and processing complexity increase
Solution Approach 1:
The capability information is segmented into common frameworks applicable to multiple network model types (reported in capability indicator field) and specific frameworks for each network model type (reported in extension fields), reducing redundant reporting.
Solution Approach 2:
The capability indicator field serves as a universal base that covers common AI frameworks supported across multiple network model types, while extension fields provide specific details, creating a multi-functional reporting structure.
3Reliability
If the capability report includes detailed information for each network model type, then the network device can make precise resource allocation decisions, but the capability indicator field exceeds maximum length
Solution Approach 1:
Detailed capability information is segmented between the capability indicator field (for common frameworks) and capability indicator extension fields (for specific frameworks of each network model type), maintaining both detail and field length constraints.
Solution Approach 2:
The extension field acts as an intermediary that carries detailed specific capability information without burdening the main capability indicator field, allowing precise resource allocation while maintaining field length limits.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
A method and apparatus for reporting an AI network model support capability, a method and apparatus for receiving an AI network model support capability, and a storage medium, a user equipment and a base station. The method for reporting an AI network model support capability comprises: determining a capability of supporting an AI network model, wherein the capability of supporting an AI network model comprises whether the AI network model being used to perform channel estimation is supported; and during a random access process, reporting, by using an uplink resource, the capability of supporting the AI network model, or triggering, in a connected state, the reporting of the capability of supporting the AI network model. By means of the technical solution of the present invention, a base station can learn of related parameters, about an AI network model, at a user equipment side.