Skip to content

ROADMAP.md

Latest commit

 

History

History
229 lines (172 loc) · 15 KB

ROADMAP.md

File metadata and controls

229 lines (172 loc) · 15 KB
Raw

6G Experiment Bed — Project Strategy & Engineering Roadmap

This is an ambitious project. Here's how to approach it with the rigor it demands.

1. Situational Awareness: Where 6G Standards Actually Are (Feb 2026)

Before writing a line of spec-driven code, internalize the standards landscape:

Body Status
3GPP Study items in Rel-19/20. No 6G NR spec frozen yet. Expected ~Rel-21 (2028)
ITU-R IMT-2030 Framework published (2023), requirements set; no air interface spec
ETSI Working groups active on AI/ML integration
O-RAN Alliance 6G use cases being defined in WG1

Implication for your project: You are building a research testbed, not a standards-compliance implementation. Your target is to model proposed architectures from academic papers + industry whitepapers (Qualcomm, Ericsson, Nokia, Samsung 6G papers), not 3GPP TSes. This is actually an advantage — you have freedom to experiment.

2. Architectural Foundation: 6G vs 5G — What You Must Understand

Since you know 5G call flows deeply, here's the delta mental model:

2.1 The 6G Stack Additions (beyond 5G)

┌─────────────────────────────────┐
│   Semantic / Goal-oriented      │  ← NEW: transmit meaning, not bits
├─────────────────────────────────┤
│   AI/ML Native Layer            │  ← NEW: protocol decisions via inference
├─────────────────────────────────┤
│   6G Core (User-plane first)    │  ← REARCHITECTED: flatter than 5GC
├─────────────────────────────────┤
│   RRC (simplified)              │  ← SIMPLIFIED vs 5G RRC complexity
├─────────────────────────────────┤
│   PDCP / RLC / MAC              │  ← EVOLVED (same concepts, new params)
├─────────────────────────────────┤
│   PHY (THz + RIS + MIMO)        │  ← RADICALLY NEW
├─────────────────────────────────┤
│   NTN (Sat + HAPS + UAV)        │  ← NEW as native (not bolted-on like 5G NTN)
├─────────────────────────────────┤
│  ISAC (Sensing + Comms unified) │  ← NEW: same waveform for radar+comms
└─────────────────────────────────┘

2.2 The Qualcomm "User-plane first" Insight

In 5G, NAS and RRC are enormous — registration, authentication, session management all flow through complex control plane signaling you know well (AMF, AUSF, SEAF). 6G proposals (Qualcomm Foundry paper series) argue for collapsing this: the user plane should drive service access, with control plane as a thin adaptation layer. This means your 6g-core crate's AMF-style entities need rethinking, not just renaming. github

3. Project Planning: Phases

Phase 0 — Foundation (Weeks 1-4)

Goal: Stable, buildable workspace + spec reading discipline

Tasks:

  • Add CI (GitHub Actions): cargo test, cargo clippy -- -D warnings, cargo fmt --check
  • Create a docs/ folder: for each crate, write a 1-page design doc before implementation (spec → design → code order)
  • Pin your reference papers in docs/references.md: ITU-R M.2160, Samsung 6G Vision (2020), Nokia Bell Labs 6G (2021), Qualcomm Foundry series, key IEEE papers on ISAC and RIS
  • Rework 6g-common/types.rs: define your fundamental types precisely — Frequency (with THz range), Position3D, BearerId, SliceId, Payload — these are your API contracts

Phase 1 — PHY Layer Experiment (Weeks 5-12)

Goal: Simulate one end-to-end waveform through the physical layer

This is your experiment bed's core value. 6G PHY is where the biggest unknowns are.

Sub-tasks:

  • Waveform module (6g-phy/waveform.rs): Implement OFDM baseline, then OTFS (Orthogonal Time Frequency Space) — the 6G candidate waveform for high-mobility scenarios. Compare delay-Doppler domain vs time-frequency (5G NR is OFDM)
  • Spectrum module (6g-phy/spectrum.rs): Model sub-THz band (100-300 GHz). Key parameter: oxygen absorption at 60 GHz, rain fade. You don't need actual RF — model path loss as PL(d) = FSPL + α·d with THz-specific α
  • MIMO module (6g-phy/mimo.rs): Implement massive MIMO channel model (3GPP-style CDL or QuaDRiGa), then extend to Extremely Large Aperture Array (ELAA) — the 6G MIMO paradigm. Near-field effects become relevant at THz
  • RIS module (6g-phy/ris.rs): Reconfigurable Intelligent Surface — model as a phase-shift matrix applied to the channel. Even a simplified H_eff = H_d + H_r * Φ * H_i model is valuable
  • Validation: Write unit tests that verify SNR vs distance curves match published results for THz channels

Phase 2 — ISAC Integration (Weeks 13-18)

Goal: One waveform, dual function — sense and communicate

ISAC is one of the clearest 6G differentiators from 5G. Your existing radar/sensing intuition from analog design applies here.

Sub-tasks:

  • Define the ISAC waveform tradeoff: CRB (Cramér-Rao Bound) for sensing vs capacity for communication — implement as a Pareto frontier computation
  • Implement DFRC (Dual Function Radar Communications) basic model: OFDM with embedded sensing sequences
  • Target detection stub: FFT-based range-Doppler processing on the reflected signal
  • Validate against: SINR for communication link, detection probability (Pd) vs false alarm (Pfa) for sensing

Phase 3 — MAC/RLC/PDCP Layer (Weeks 19-26)

Goal: Functional data plane from application to PHY

Your 5G experience is most transferable here. Key 6G differences:

  • AI-native scheduler in MAC: implement a simple Q-learning or bandit-based scheduler alongside a classic Round Robin — compare. This is the experiment
  • HARQ evolution: 6G proposes proactive HARQ (predictive retransmission). Implement classic Chase Combining first, then model a prediction oracle
  • PDCP: Fill in the process_tx/process_rx TODOs — implement actual ROHC simulation (even simplified), proper sequence numbering, and replay detection
  • RLC: AM/UM/TM modes — directly map your 5G RLC knowledge

Phase 4 — Core Network (Weeks 27-36)

Goal: Minimal 6G control plane that differs meaningfully from 5GC

  • Re-architect 6g-core away from direct 5G NF mapping. Design a Service-Based Architecture v2 with flatter hierarchy
  • Implement a registration flow that is simpler than 5G Registration (no AMF reselection complexity, no full NAS Security Mode Command chain) — this is the research hypothesis
  • Digital Twin integration stub: the network maintains a real-time model of its own state — implement as a simple state-snapshot + diff mechanism
  • NTN handover procedure: LEO satellite → terrestrial handover, accounting for the ~1.8ms propagation delay already modeled in the skeleton
  • GnbNode in 6g-core: wires RrcLayer + PdcpEntity to N2 (AMF) and N3 (UPF) interface stubs, completing the RAN ↔ Core bridge

Phase 5 — Semantic & AI Layers (Weeks 37-48)

Goal: Demonstrate semantic communication on one end-to-end flow

These are the most speculative but highest-impact layers:

  • Semantic encoder/decoder: TextSemanticCodec — deterministic term-frequency encoder (64-byte signature, ~15× compression). ONNX-based sentence transformer is tracked as Phase 6+ work.
  • AI inference crate (6g-ai): Implement a channel estimation neural network (simple MLP) — compare with LS/MMSE estimators at various SNRs
  • Goal-oriented communication: Define a task (e.g., "transmit enough for image classification to succeed") — measure task success rate, not BER

Phase 6 — Full 6G Core Architectural Differentiators

Goal: Make the core genuinely 6G, not just a better 5G SA

  • Semantic PDU sessions (6g-core/smf.rs, upf.rs): PduSessionType::Semantic(GoalSpec) — the 4th PDU session type unique to 6G; UPF routes via TextSemanticCodec
  • User-plane-first (6g-core/upf.rs): UPF::forward_unknown_flow() → FlowAction; lazy session establishment; control plane is a thin adaptation layer
  • SDF — Sensing Data Function (6g-core/sdf.rs): 6G-new NF with no 5G equivalent; exposes ISAC RAN sensing results as a core SBI subscription service
  • NTN-aware AMF (6g-core/amf.rs): TrackingArea::Ntn { ntn_node_id, beam_id, propagation_delay: Duration } — first-class NTN mobility in the AMF
  • NRF capability graph (6g-core/nrf.rs): NfCapability enum + discover_by_capability() — replaces the 5G static endpoint table with knowledge-graph-style discovery
  • Full teardown path: CoreNetwork::deregister_ue(), release_session(), per-session UPF bearer stats
  • AUSF / UDM: Subscriber credential store + 5G-AKA conceptual auth flow
  • NSSF admission control: load-based per-slice rejection (admit_ue, release_ue)
  • ONNX-based sentence transformer codec (OnnxSemanticCodec + OnnxModel): 32-byte quantised embedding, 31.25× compression, semantic similarity preserved; swap OnnxModel for ort::Session to use a real .onnx file (see docs/6g-ai.md)

4. Validation Strategy

For an experiment bed, validation = reproducible, measurable results against known baselines.

Layer Baseline (5G/Theory) 6G Experiment Metric
PHY waveform OFDM in AWGN OTFS in delay-Doppler BER vs Eb/N0
MIMO 8x8 spatial multiplexing ELAA near-field Spectral efficiency (bps/Hz)
RIS No RIS (direct link) RIS-assisted link Coverage extension in dB
ISAC Separate radar + comms DFRC waveform CRB (sensing) + Rate (comms)
MAC scheduler Round Robin AI Q-learning Throughput fairness (Jain index)
Semantic comms Raw bit transmission Semantic encoding (TextSemanticCodec) Task success rate vs bandwidth
Core — teardown 5G SA PDU teardown deregister_ue() / release_session() State consistency
Core — user-plane-first Pre-established session forward_unknown_flow() lazy setup First-packet latency
Core — sensing Not exposed in 5G core SDF pub/sub Subscription delivery rate
NTN handover Terrestrial-only LEO-assisted handover (TrackingArea::Ntn) Handover latency

Every experiment needs: input parameters → deterministic simulation → output metrics → comparison against baseline. Use Rust's criterion crate for micro-benchmarks and write a scripts/ folder with Python/R notebooks to plot results.

5. Specification Guidance — What to Read and When

Now (before Phase 1):

  • ITU-R M.2160 — IMT-2030 Framework (free download) — the requirements document
  • Samsung 6G Vision white paper (2020) — best overview of use case taxonomy
  • Qualcomm 6G Foundry series (5 papers, all free on qualcomm.com) — especially "Rethinking the Control Plane" and "AI-native 6G" github

During PHY Phase:

  • IEEE papers: "OTFS Modulation" (Hadani et al., 2017), RIS channel modeling (Basar et al., 2019)
  • 3GPP TR 38.901 (channel models) — your 5G NR channel model reference, directly applicable
  • IEEE ComSoc 6G roadmap papers (2021-2023)

During ISAC Phase:

  • Liu, F. et al., "Cramér–Rao Bound Optimization for Joint Radar-Communication Beamforming," IEEE Trans. Signal Process. 2018, DOI: 10.1109/TSP.2018.2864261
  • 3GPP TR 22.837 — ISAC study item (even though 5G-era, defines sensing use cases)

During Core Network Phase:

  • ETSI ENI (Experiential Networked Intelligence) specs
  • O-RAN Alliance: "AI/ML Workflow Description and Requirements" (O-RAN.WG2.AIML)

Avoid reading 3GPP 6G work items as normative specs — they are incomplete study items and will cause premature spec-anchoring.

6. Reasoning Depth for Each Dev Task

Apply this decision framework consistently:

For every module:
1. What does the 5G equivalent do? (your baseline)
2. What does the 6G proposal change and WHY? (the research question)
3. What is the simplest model that captures the change? (MVP)
4. What does success look like, quantitatively? (the test)
5. What are the known failure modes? (risk)

Example applied to RIS (6g-phy/ris.rs):

  1. 5G: passive reflectors don't exist — channel is given
  2. 6G: RIS actively phase-shifts reflected paths to constructively combine at receiver — extends coverage without power amplifiers
  3. MVP: H_eff = H_direct + H_reflect * Φ * H_incident where Φ is a diagonal phase matrix you optimize
  4. Success: received SNR with RIS > without RIS by > 10 dB in a shadowed scenario
  5. Risk: near-field vs far-field assumption — at THz, UE might be in near-field of the RIS, invalidating the far-field channel model

7. Comparing Against Real Systems

The experiment bed is only trustworthy when its outputs can be verified against systems that are already built. The comparison methodology — which simulators to run, which public datasets to download, and how to import external results into the ValidationCheck framework — is documented in docs/comparison-strategy.md.

The code-level hook is sixg_common::baseline::BaselineDataset:

let dataset = BaselineDataset::from_csv_str(csv_str, source)?;
let result = dataset.compare(|snr_db| simulate_ber(snr_db), 5.0);
assert!(result.passed(), "{}", result.summary());

Key real-system targets, in priority order:

Phase Real system Metric to match
PHY NIST 28 GHz path-loss tables Path loss (dB) vs distance
PHY Vienna 5G LLS BER vs Eb/N0 for OTFS
MAC ns-3 NR (5G-LENA) Jain fairness index at 20 UEs
MAC OAI 5G SA traces HARQ BLER vs SNR
ISAC Liu et al. Table II (IEEE TSP 2018) CRB vs sensing power ratio
System srsRAN 5G SA End-to-end throughput / latency

Given this is a solo research project that you might publish or use for PhD work:

  • One crate, one responsibility: Don't let 6g-core absorb PHY concerns
  • Feature flags: Use cargo features to gate speculative features (#[cfg(feature = "semantic")]) — keeps the base always compilable
  • Trait-based interfaces: Define trait Encoder, trait Scheduler, trait ChannelModel — lets you swap implementations without restructuring
  • Reproducible experiments: Seed all RNGs, record all parameters in JSON output files alongside results
  • GitHub Issues = experiment backlog: One issue per experiment, not per task. Title format: [EXP] ISAC: DFRC waveform vs separate radar+comms at 100 GHz