AI-Based CSI Feedback for Massive MIMO Channel Reporting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge in MIMO technology is the high amount of channel state information (CSI) feedback required in massive MIMO scenarios, which consumes valuable uplink resources, and there is a need for more efficient CSI feedback methods.
Innovation Solution
Implementing AI/ML-based CSI generation and reconstruction processes at both the terminal equipment and network device sides, utilizing AI/ML models to generate and process multiple types of CSI based on identical CSI-RSs, allowing the network device to monitor and adjust the CSI generation model performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional CSI feedback methods are used in massive MIMO scenarios, then downlink channel state information can be obtained for precoding, but the amount of uplink feedback resources consumed becomes very large
Solution Approach 1:
The patent segments the CSI feedback process into two parts: the terminal equipment generates multiple types of CSI (including AI/ML-based CSI and traditional codebook-based CSI) and feeds them back separately, allowing the network device to select appropriate feedback methods for different scenarios. This segmentation enables flexible resource allocation and reduces overall feedback overhead while maintaining accuracy.
Solution Approach 2:
The patent introduces AI/ML-based CSI generation as a new parameter/method alongside traditional codebook-based methods. The network device can dynamically switch between different CSI generation approaches (AI/ML model parameters vs. codebook parameters) based on channel conditions and resource availability, optimizing the balance between feedback amount and accuracy.
2Quantity of substance
If AI/ML-based CSI generation is implemented, then CSI feedback amount is reduced and accuracy is enhanced, but system complexity increases due to model training and inference requirements
Solution Approach 1:
The patent applies preliminary action by pre-training AI/ML models offline before deployment. The model training and parameter optimization are performed in advance, and only the trained model parameters are deployed in the terminal equipment. This separates the complex training process from real-time operation, reducing runtime complexity while maintaining the benefits of AI/ML-based CSI generation.
Solution Approach 2:
The patent introduces an intermediary layer (the AI/ML model) between channel measurement and CSI generation. Instead of directly generating CSI from channel measurements using traditional methods, the AI/ML model acts as an intermediary that processes channel information and generates optimized CSI, reducing feedback overhead while improving accuracy.
3Adaptability or versatility
If multiple types of CSI are fed back from terminal equipment, then network device can monitor model performance and switch between methods, but terminal processing complexity increases
Solution Approach 1:
The patent implements dynamics by enabling the terminal equipment to dynamically generate different types of CSI (AI/ML-based and codebook-based) based on network device instructions and current channel conditions. The terminal can adaptively switch between generation methods and adjust feedback content, providing flexibility while distributing processing complexity according to capability and need.
Data Source
AI summary
An apparatus for receiving channel state information (CSI), applicable to a network device, includes: first processor circuitry making the network device execute the following: transmitting information by the network device to a terminal equipment, at least a part of the information being one or more CSI reporting configurations, the one or more CSI reporting configurations being used for indicating the terminal equipment to transmit at least two of first channel state information (CSI), second CSI and third CSI; and receiving, by the network device, at least two of the first CSI, the second CSI and the third CSI transmitted by the terminal equipment, wherein at least two of the first CSI, the second CSI and the third CSI are generated based on at least a part of identical channel state information reference signals (CSI-RSs) and/or CSI-RS resources.