Integrated Dual-Sided Machine Learning for Efficient CSI Acquisition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wireless communication systems face challenges in accurately acquiring channel state information (CSI) with high efficiency, leading to increased resource occupation and overhead due to the large amount of information required for precise channel representation, which can be addressed through dual-sided machine learning models integrated between user equipment (UE) and network.
Innovation Solution
The integration of dual-sided machine learning models between UE and network, optimized for processing capabilities and storage space, allows for high-accuracy CSI acquisition while minimizing resource usage, facilitated by a common server for model and data set sharing and management, and selection of appropriate models based on UE capabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional CSI reporting procedures are used to accurately represent channel information, then measurement precision is improved, but device complexity and resource occupation increase due to large information overhead
Solution Approach 1:
The patent transforms the CSI reporting problem by changing the parameter representation from traditional explicit channel state parameters to implicit neural network model parameters. Instead of transmitting detailed channel matrices, the system transmits compact neural network weights and biases that can reconstruct channel information, significantly reducing overhead while maintaining accuracy
Solution Approach 2:
The patent uses neural networks to create a compressed representation (copy) of the channel state information. The neural network model learns to map channel inputs to CSI outputs, creating a compact surrogate that replicates the function of traditional CSI reporting without requiring transmission of the full channel matrix
2Productivity
If dual-sided machine learning models are integrated between UE and network, then productivity is improved through efficient CSI acquisition, but device complexity increases due to model integration requirements
Solution Approach 1:
The patent divides the machine learning system into two separate but coordinated components: a UE-side neural network for channel estimation and a network-side neural network for CSI generation. Each side operates independently with its own model, avoiding the complexity of fully integrated models while achieving collaborative intelligence
Solution Approach 2:
The patent introduces a centralized server as an intermediary that manages the training, updates, and synchronization of neural network models for both UE and network. This intermediary handles the complexity of model integration, allowing the actual inference processes at UE and network to remain relatively simple
3Adaptability or versatility
If machine learning models are pre-installed or exchanged between UE and network, then adaptability is improved for dual-sided learning, but loss of time occurs during model installation and synchronization
Solution Approach 1:
The patent performs model training and preparation in advance using a centralized server. Neural network models are pre-trained on historical channel data and prepared for deployment before actual CSI acquisition needs arise. This preliminary action ensures models are ready when needed, eliminating runtime training delays
Solution Approach 2:
The patent designs a universal model management architecture that can handle multiple learning scenarios (fully distributed, centralized, semi-centralized) through a single flexible framework. The same infrastructure supports different model exchange and synchronization strategies, reducing the need for separate implementation paths
Data Source
AI summary
The present disclosure generally relates to wireless communication systems, and more particularly, to an apparatus and method for integrated inference using dual-sided machine learning in wireless communication systems. A method of operating a user equipment (UE) in a wireless communication system includes: transmitting capability information of the UE to a network; receiving at least one of a structure or parameters of a reference model, or receiving a learning data set from the network according to the capability information of the UE; configuring a machine learning (ML) model directly on the UE or through a UE-side learning server based 10 on the received information; and performing integrated inference based on dual-sided machine learning models with the network using the configured machine learning model.