R1-2600148
discussion
Transmission schemes for 6GR DL shared channels
From Huawei
Summary
This document presents Huawei's views on 6G downlink transmission enhancements, containing 24 Observations and 24 Proposals across requirements, transmission schemes, MIMO layers, resource allocation, multi-TRP design, DMRS design (including AI/ML), and evaluation methodology.
Position
Huawei proposes that 6GR DL transmission adopt a DMRS-based transparent scheme as baseline, supporting high-order SU-MIMO up to 16 layers and MU-MIMO up to ~100 layers to meet 3x average spectral efficiency targets. They require single-sided (UE-sided) AI/ML receiver models as the starting point for DMRS overhead reduction, prioritizing sparse orthogonal DMRS (sub-use case A) over superimposition (SIP) or DMRS-free schemes due to scheduling restrictions, UE complexity, latency, and generalization concerns with the latter. They propose utilizing long-term channel information as assistance information for both non-AI and AI/ML DMRS channel estimation, showing up to 111% SE gain over NR ideal Wiener filtering with sparse DMRS. They propose studying DMRS expansion to 96 orthogonal ports using frequency-/code-domain multiplexing with NR DMRS Type 2 as a design starting point, while requiring scalable DMRS port design and unified AI/non-AI DMRS patterns to guarantee flexible MU-MIMO coexistence. For mTRP, they propose incorporating CJT as the primary scheme from the initial release with coordination sets up to 9 TRPs, and for inter-layer SINR imbalance, they propose finer CW-to-layer mapping granularity showing 24% average spectral efficiency gain.
Key proposals
- Proposal 1 (Sec 2.1): High-order MIMO transmission and larger bandwidth data transmission for 3x average spectral efficiency and ~2x peak data rate; advanced mTRP transmission for 3x 5%-tile user spectral efficiency improvement.
- Proposal 2 (Sec 2.2.1): Leverage larger MU-MIMO potential from better inter-UE spatial orthogonality and larger MIMO transmit diversity potential from larger bandwidth frequency selectivity for 7GHz bands.
- Proposal 4 (Sec 3.1.1): Adopt DMRS-based transparent scheme as baseline DL transmission scheme for 6GR.
- Proposal 6 (Sec 3.2): Support high-order SU-MIMO transmission up to 16 layers and high-order MU-MIMO transmission up to ~100 layers.
- Proposal 8 (Sec 3.3): Study FDRA mechanism including RBG-based and RIV-based types considering wider bandwidth, BWP/carrier switch, and SBFD.
- Proposal 11 (Sec 3.4): Study solutions to improve overall system throughput for scenarios with significant SINR imbalance among MIMO layers.
- Proposal 12 (Sec 4.2): Study mTRP transmission schemes (CJT, NCJT, DPS) with enhancements including larger number of coordinated TRPs up to 9.
- Proposal 13 (Sec 5.1.1): Study and consider larger number of DMRS orthogonal ports up to 96, low overhead DMRS, and high-accuracy channel estimation for low overhead DMRS design.
- Proposal 14 (Sec 5.1.2): Strive for unified AI and Non-AI DMRS design, scalable DMRS design, and use NR DMRS Type 2 as starting point.
- Proposal 15 (Sec 5.2.1): Study and consider utilization of long-term channel information for low overhead DMRS design.
- Proposal 17 (Sec 5.2.2): Study and consider PDSCH DMRS sequence design for reduced inter-sequence interference.
- Proposal 19 (Sec 5.2.4): For low overhead DMRS with AI/ML receiver, study single-sided model (UE-sided model for DL) as starting point.
- Proposal 21 (Sec 5.2.4): Study sub-use case A (sparse DMRS) as starting point for AI/ML-based DMRS overhead reduction to enable flexible co-scheduling with non-AI/ML UEs.
- Proposal 22 (Sec 5.2.5): Study LCM aspects including label acquisition methods, inference-related aspects, and UE-side monitoring procedure for AI/ML DMRS.
- Proposal 24 (Sec 6.2): Consider AI/ML evaluation methodology including KPIs (BLER, spectral efficiency, FLOPs/MACs, number of parameters), generalization/scalability performance, and imperfect/non-ideal factors.