TR 26.870 V0.2.12 – Study on Media Aspects for 6G System (Release 20)

Overview
This is an early-stage Technical Report (SA4 study item FS_6G_MED) identifying media-related opportunities, gaps, and requirements for 6G. It is not yet approved and has no normative status. Conclusions will feed future detailed studies and normative work.

Scope and Alignment
- Aligned with SA1 service requirements (TR 22.870), SA2 architecture study (TR 23.801-01), and existing SA4 5G media specs (TS 26.501, TS 26.506, TS 26.511, TS 26.512, TS 26.114, etc.)
- Existing 5G media delivery architecture (streaming and real-time communication) is the baseline starting point

Five Work Topics Defined (all still under development)
1. WT#1 – Media Delivery Architecture: Harmonized 6G media delivery architecture building on TS 26.501/26.506; key issues include whether streaming and real-time communication architectures should be harmonized or separated, and which content delivery protocols to adopt/extend for 6G
2. WT#2 – 6G Media: Immersive/AI media trends, traffic characterization, QoE metrics, multimodal formats, AI-related media services (agents, LLMs, diffusion models), and 2D video codec updates (H.264/HEVC successors)
3. WT#3 – Media Aspects Related to SA2 Topics: Placeholder; content TBD
4. WT#4 – Media for Ubiquitous Access: Media over NTN and low-bitrate/low-power scenarios; references ULBC study (TR 26.940); key issues include NTN bitrate/latency/loss ranges and performance of existing services under such conditions
5. WT#5 – Trusted and Private Communication for Media: Placeholder; content TBD

Key Technical Observations (Annex A, partially agreed)
- AI-enabled applications generate heterogeneous, multimodal, uplink-heavy, bursty traffic with diverse latency sensitivity
- Current QoS frameworks may lack granularity and context-awareness for AI-driven traffic
- Multi-device AI service scenarios are not well addressed by existing UE-centric assumptions
- Embodied AI (robots, UAVs) may require 20–100 Mbit/s uplink with low latency and error resilience
- Temporal synchronization across modalities (video, audio, haptics, AI data) is a key challenge

Tooling Contributions (Annexes B & C)
- Network Emulator (Annex B): Linux tc/netem-based emulator with YAML-configured profiles; includes 5QI-derived profiles (5QI 1/2/7/80), scenario profiles (ideal 6G, 5G urban, satellite, cell edge, congested), and advanced impairment models (Gilbert-Elliott loss, HTB shaping). Hosted at GitHub (5G-MAG/6G-Testbed)
- AI Traffic Characterization Testbed (Annex C): End-to-end orchestration framework for AI service traffic measurement; supports OpenAI/Gemini/DeepSeek/vLLM providers; captures L3/L4 (pcap) and L7 (mitmproxy); logs TTFT, TTLT, token rate, burstiness, UL/DL bytes to SQLite. Use of TTFT/TTLT for AI traffic evaluation is FFS.

Status
Document is v0.2.12; most clause content remains under editor's notes. Findings (clause 7) and recommendations (clause 8) are empty pending completion of work topics.