Joint CSI Feedback for Cross-Vendor AI/ML Model Training

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The challenge in wireless communication systems is the scalability and compatibility of AI/ML models for channel state information feedback when UE-side and network-side models are controlled by different vendors, requiring extensive collaboration that is not scalable.

Innovation Solution

Implementing joint CSI feedback that includes a known mapping and a UE-side AI/ML model, allowing the network node to interpret and train a network-side AI/ML model using over-the-air feedback, without requiring extensive vendor collaboration.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If extensive vendor collaboration is required for AI/ML model compatibility, then model compatibility can be achieved, but scalability deteriorates

Engineering Contradiction:

Improvemodel compatibilityVSAvoidscalability

Core Design Contradiction:

ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent introduces a standardized interface as an intermediary between UE-side AI/ML models and network-side AI/ML models. This interface acts as a mediator that enables compatibility without requiring direct vendor collaboration, allowing models from different vendors to work together through the standardized mapping layer.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent segments the AI/ML model into two independent parts: a UE-side AI/ML model that processes channel state information locally, and a network-side AI/ML model that performs downstream processing. These segmented models communicate through a standardized interface, enabling independent development by different vendors while maintaining compatibility.

Inventive Principle:

Principle #1Segmentation

2Productivity

If UE-side AI/ML model is used for CSI feedback, then system performance improves, but device complexity increases

Engineering Contradiction:

Improvesystem performanceVSAvoiddevice complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent performs preliminary action by training the UE-side AI/ML model beforehand to map channel state information to a standardized interface format. This pre-training enables the UE to autonomously prepare feedback data in the required format without requiring complex real-time processing or collaboration during operation.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20250261022A1Reporting channel state information feedback for network-side model training

Publication Date: 2025.08.14 QUALCOMM INC

AI summary

Various aspects of the present disclosure generally relate to wireless communication. In some aspects, a user equipment (UE) may receive a configuration associated with artificial intelligence or machine learning (AI/ML) model based channel state feedback (CSF) reporting. The UE may transmit, based at least in part on the configuration, a channel state information (CSI) report that indicates joint CSI feedback, wherein the joint CSI feedback includes a first CSI associated with a known mapping and a second CSI associated with a UE-side AI/ML model. Numerous other aspects are described.