AI for CSI Feedback Enhancement in 5G-Advanced

TL;DR

An overview of AI for CSI feedback enhancement in 5G-Advanced and the main challenges and open problems in the standardization of AI for CSI feedback enhancement are identified and discussed.

Abstract

The 3rd Generation Partnership Project started the study of Release 18 in 2021. Artificial intelligence (AI)-native air interface is one of the key features of Release 18, where AI for channel state information (CSI) feedback enhancement is selected as the representative use case. This article provides an overview of AI for CSI feedback enhancement in 5G-Advanced. Several representative non-AI and AI-enabled CSI feedback frameworks are first introduced and compared. Then, the standardization of AI for CSI feedback enhancement in 5G-advanced is presented in detail. First, the scope of the AI for CSI feedback enhancement in 5G-Advanced is presented and discussed. Then, the main challenges and open problems in the standardization of AI for CSI feedback enhancement, especially focusing on performance evaluation and the design of new protocols for AI-enabled CSI feedback, are identified and discussed. This article provides a guideline for the standardization study of AI-based CSI feedback enhancement.

Figures (4)

• Figure 1: Three representative frameworks of AI for CSI feedback, including (a) one-sided refinement for implicit CSI feedback, (b) two-sided enhancement for implicit CSI feedback, and (c) two-sided enhancement for explicit CSI feedback.
• Figure 2: Illustration of CSI compression and reconstruction with the aid of learned environmental knowledge
• Figure 3: Three collaborations between the UE and the BS of the AI-enabled CSI feedback.
• Figure 4: Three frameworks for joint AI-enabled CSI feedback and prediction. The first two NN-based prediction modules work at the UE and the BS, and the last one implicitly realizes channel prediction.