R1-2600152
discussion
Aspects of uplink-based CSI acquisition
From Huawei
Summary
This document from Huawei presents 15 proposals and 7 observations on 6GR uplink-based CSI acquisition via SRS, addressing challenges like low SNR, channel aging, and inter-cell interference through sparse patterns, AI/ML, long-term channel information, and evaluation methodologies.
Position
Huawei proposes studying sparse SRS design using larger transmission comb values (e.g., Comb-8) and larger partial frequency factors to improve SNR and reduce channel aging. They propose studying long-term channel information leveraging to assist CSI acquisition under sparse SRS patterns, reporting 24% to 52% spectral efficiency gain. For AI/ML, they prioritize sub-use case A (low overhead SRS with AI/ML) and require a unified SRS pattern/sequence design enabling FDM/CDM coexistence between AI/ML and non-AI/ML SRS patterns. They propose studying DMRS measurement assistance for AI/ML-based NW-side channel estimation to alleviate channel aging. They specify evaluation methodologies including SGCS, NMSE, BLER, spectral efficiency KPIs and FLOPs/MACs for complexity, and require generalization testing across deployment scenarios, antenna configurations, and Doppler conditions.
Key proposals
- Proposal 1 (Sec 2): 6GR shall study and overcome challenges for SRS including low SNR, channel aging, and inter-cell interference.
- Proposal 2 (Sec 3.1): 6GR shall study sparse SRS design to improve SRS SNR, increase SRS capacity so as to alleviate channel aging and/or avoid SRS interference.
- Proposal 3 (Sec 3.1.1): 6GR SRS should at least study comb-based sparse SRS pattern.
- Proposal 5 (Sec 3.1.2): 6GR should study and consider leveraging long-term channel information to facilitate high-precision CSI acquisition under sparse SRS pattern.
- Proposal 6 (Sec 3.2): 6GR shall study SRS frequency hopping for supporting larger BW to achieve SRS SNR improvement.
- Proposal 7 (Sec 3.3): 6GR should study SRS sequence design considering lower PAPR and lower cross-correlation.
- Proposal 9 (Sec 3.4): 6GR should study SRS interference randomization.
- Proposal 11 (Sec 3.5.2): 6GR shall study dynamic SRS configuration update mechanism.
- Proposal 12 (Sec 4): For the use case of AI/ML for SRS, study the sub-use case A: low overhead SRS with AI/ML with higher priority.
- Proposal 13 (Sec 4.2): For the LCM aspects of low overhead SRS with AI/ML, study the label acquisition, e.g., dense SRS pattern in spatial/frequency/temporal domain.