Intelligent native channel state information acquisition and standardization for 6G

doi:10.12267/j.issn.2096-5931.2026.10.004

Abstract

Abstract:

The deep integration of artificial intelligence and communications has emerged as one of the core driving forces advancing wireless communication systems. This paper analyzes the background and motivation of low-overhead channel state information (CSI) acquisition in 6G systems, and proposes low-overhead CSI acquisition methods for 6G systems accordingly, encompassing two categories of schemes: CSI reconstruction based on sparse reference signals and multi-scenario CSI prediction based on a single reference signal. The effectiveness of the proposed schemes is validated through simulations, and their prospects for standardization, along with the critical issues that remain to be addressed, are further discussed.

Key words: deep learning, low-overhead reference signal, channel state information reconstruction, multi-scenario channel state information prediction

CLC Number:

TN929.5

WANG Jiayin, ZUO Jun, LI Qianrui, HUANG Qiuping, FEI Yongqiang, GAO Qiubin, SUN Shaohui. Intelligent native channel state information acquisition and standardization for 6G[J]. Information and Communications Technology and Policy, 2026, 52(10): 26-39.

Add to citation manager EndNote|Ris|BibTeX

URL:

http://ictp.caict.ac.cn/EN/10.12267/j.issn.2096-5931.2026.10.004

http://ictp.caict.ac.cn/EN/Y2026/V52/I10/26

Figures/Tables 18

References 15

[1]	ITU-R. Framework and overall objectives of the future development of IMT for 2030 and beyond: recommendation ITU-R M.2160-0[R]. 2023.
[2]	ITU-R. Minimum requirements related to technical performance for IMT-2030 radio interface(s): report ITU-R M[R]. 2026.
[3]	中信科移动通信技术股份有限公司. AI与6G网络融合白皮书[R]. 2025.
[4]	3GPP. Study on enhancement for data collection for NR and EN-DC (Release 17): TR 37.817[R]. 2022.
[5]	3GPP. Study on artificial intelligence (AI)/machine learning (ML) for NR air interface (Release 18): TR 38.843[R]. 2023.
[6]	3GPP. RP-234039, new WID on artificial intelligence (AI)/machine learning (ML) for NR air interface[R]. 2023.
[7]	3GPP. RP-253340, Revised WID on AI/ML for NR air interface phase 2[R]. 2025.
[8]	3GPP. Chairman notes RAN1#123[R]. 2025.
[9]	Zhang K, Huang C, Li J L, et al. An enhanced loss function for 6G space-time domain predictive channel models[C]// 2024 IEEE 24th International Conference on Communication Technology (ICCT). Piscataway: IEEE, 2024.
[10]	Zhou T, Zhang H T, Ai B, et al. Deep-learning-based spatial-temporal channel prediction for smart high-speed railway communication networks[J]. IEEE Transactions on Wireless Communications, 2022, 21(7):5333-5345. doi: 10.1109/TWC.2021.3139384 URL
[11]	3GPP. RP-213554, new SI: study on network energy savings for NR[R]. 2021.
[12]	Wen C K, Shi W T, Jin S. Deep learning for massive MIMO CSI feedback[J]. IEEE Wireless Communications Letters, 2018, 7(5):748-751. doi: 10.1109/LWC.2018.2818160 URL
[13]	Guo J, Wen C K, Jin S, et al. Overview of deep learning-based CSI feedback in massive MIMO systems[J]. IEEE Transactions on Communications, 2022, 70(12): 8017-8045. doi: 10.1109/TCOMM.2022.3217777 URL
[14]	3GPP. Study on channel model for frequencies from 0.5 to 100 GHz: TR 38.901[R]. 2015.
[15]	3GPP. R1-2508811, moderator summary #4 on AI/ML for 6GR[R]. 2025.

参数	数值
场景	密集城区
波形	CP-OFDM
频率/GHz	4
信道模型	UMa参照TR 38.901^[14]
基站间距/m	200
时隙/非时隙	每时隙14个OFDM符号
子载波间隔/kHz	15
带宽/MHz	10
信道估计	真实估计
基站天线配置 (M, N, P, Mg, Ng; Mp, Np)	配置1:(8,8,2,1,1;4,4) 配置2:(4,4,2,1,1;4,4) 配置3:(16,16,2,1,1;4,16) 配置4:(32,16,2,1,1;8,16)
基站天线高度/m	25
基站发射功率/dBm	41
用户设备天线配置 (M, N, P, Mg, Ng; Mp, Np)	(1, 2, 2, 1, 1;1,2)
用户设备天线高度/m	1.5
用户分布	80%室内(3 km/h),20%室外(30 km/h)
每TRP用户数/个	10

参数	数值
场景	密集城区
波形	CP-OFDM
频率/GHz	4
信道模型	UMa参照TR 38.901^[14]
基站间距/m	200
时隙/非时隙	每时隙14个OFDM符号
子载波间隔/kHz	15
带宽/MHz	10
信道估计	真实估计
基站天线配置 (M, N, P, Mg, Ng; Mp, Np)	配置1:(8,8,2,1,1;4,4) 配置2:(4,4,2,1,1;4,4) 配置3:(16,16,2,1,1;4,16) 配置4:(32,16,2,1,1;8,16)
基站天线高度/m	25
基站发射功率/dBm	41
用户设备天线配置 (M, N, P, Mg, Ng; Mp, Np)	(1, 2, 2, 1, 1;1,2)
用户设备天线高度/m	1.5
用户分布	80%室内(3 km/h),20%室外(30 km/h)
每TRP用户数/个	10

参数	数值
训练方法	离线训练
损失函数	均方误差
数据规模/k	63
训练数据集规模/k	56.7
验证数据集规模/k	3.15

参数	数值
训练方法	离线训练
损失函数	均方误差
数据规模/k	63
训练数据集规模/k	56.7
验证数据集规模/k	3.15

稀疏图样	输入端口数	输出端口数	频域资源块数/个	SGCS(AI)	SGCS(采样保持)
空域稀疏	32	128	48	0.971	0.161
频域稀疏	128	128	12	0.967	0.688