Thanks to visit codestin.com
Credit goes to github.com

Skip to content

Commit 38eb93e

Browse files
ruvnetruvnet
authored
Merge pull request ruvnet#69 from ruvnet/adr-027-cross-environment-domain-generalization
feat: ADR-027 MERIDIAN — Cross-Environment Domain Generalization
2 parents d8fd5f4 + eab364b commit 38eb93e

23 files changed

Lines changed: 2575 additions & 39 deletions

‎CHANGELOG.md‎

Lines changed: 8 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -8,6 +8,14 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
88
## [Unreleased]
99

1010
### Added
11+
- **Project MERIDIAN (ADR-027)** — Cross-environment domain generalization for WiFi pose estimation (1,858 lines, 72 tests)
12+
- `HardwareNormalizer` — Catmull-Rom cubic interpolation resamples any hardware CSI to canonical 56 subcarriers; z-score + phase sanitization
13+
- `DomainFactorizer` + `GradientReversalLayer` — adversarial disentanglement of pose-relevant vs environment-specific features
14+
- `GeometryEncoder` + `FilmLayer` — Fourier positional encoding + DeepSets + FiLM for zero-shot deployment given AP positions
15+
- `VirtualDomainAugmentor` — synthetic environment diversity (room scale, wall material, scatterers, noise) for 4x training augmentation
16+
- `RapidAdaptation` — 10-second unsupervised calibration via contrastive test-time training + LoRA adapters
17+
- `CrossDomainEvaluator` — 6-metric evaluation protocol (MPJPE in-domain/cross-domain/few-shot/cross-hardware, domain gap ratio, adaptation speedup)
18+
- ADR-027: Cross-Environment Domain Generalization — 10 SOTA citations (PerceptAlign, X-Fi ICLR 2025, AM-FM, DGSense, CVPR 2024)
1119
- **Cross-platform RSSI adapters** — macOS CoreWLAN (`MacosCoreWlanScanner`) and Linux `iw` (`LinuxIwScanner`) Rust adapters with `#[cfg(target_os)]` gating
1220
- macOS CoreWLAN Python sensing adapter with Swift helper (`mac_wifi.swift`)
1321
- macOS synthetic BSSID generation (FNV-1a hash) for Sonoma 14.4+ BSSID redaction

‎README.md‎

Lines changed: 4 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -49,7 +49,7 @@ docker run -p 3000:3000 ruvnet/wifi-densepose:latest
4949
| [User Guide](docs/user-guide.md) | Step-by-step guide: installation, first run, API usage, hardware setup, training |
5050
| [WiFi-Mat User Guide](docs/wifi-mat-user-guide.md) | Disaster response module: search & rescue, START triage |
5151
| [Build Guide](docs/build-guide.md) | Building from source (Rust and Python) |
52-
| [Architecture Decisions](docs/adr/) | 26 ADRs covering signal processing, training, hardware, security |
52+
| [Architecture Decisions](docs/adr/) | 27 ADRs covering signal processing, training, hardware, security, domain generalization |
5353

5454
---
5555

@@ -75,6 +75,7 @@ The system learns on its own and gets smarter over time — no hand-tuning, no l
7575
|---|---------|---------------|
7676
| 🧠 | **Self-Learning** | Teaches itself from raw WiFi data — no labeled training sets, no cameras needed to bootstrap ([ADR-024](#self-learning-wifi-ai-adr-024)) |
7777
| 🎯 | **AI Signal Processing** | Attention networks, graph algorithms, and smart compression replace hand-tuned thresholds — adapts to each room automatically ([RuVector](#ai-backbone-ruvector)) |
78+
| 🌍 | **Works Everywhere** | Train once, deploy in any room — adversarial domain generalization strips environment bias so models transfer across rooms, buildings, and hardware ([ADR-027](#cross-environment-generalization-adr-027)) |
7879

7980
### Performance & Deployment
8081

@@ -107,7 +108,7 @@ Neural Network maps processed signals → 17 body keypoints + vital signs
107108
Output: real-time pose, breathing rate, heart rate, presence, room fingerprint
108109
```
109110

110-
No training cameras required — the [Self-Learning system (ADR-024)](#self-learning-wifi-ai-adr-024) bootstraps from raw WiFi data alone.
111+
No training cameras required — the [Self-Learning system (ADR-024)](#self-learning-wifi-ai-adr-024) bootstraps from raw WiFi data alone. [MERIDIAN (ADR-027)](#cross-environment-generalization-adr-027) ensures the model works in any room, not just the one it trained in.
111112

112113
---
113114

@@ -511,6 +512,7 @@ The neural pipeline uses a graph transformer with cross-attention to map CSI fea
511512
| [RuVector Crates](#ruvector-crates) | 11 vendored Rust crates from [ruvector](https://github.com/ruvnet/ruvector): attention, min-cut, solver, GNN, HNSW, temporal compression, sparse inference | [GitHub](https://github.com/ruvnet/ruvector) · [Source](vendor/ruvector/) |
512513
| [AI Backbone (RuVector)](#ai-backbone-ruvector) | 5 AI capabilities replacing hand-tuned thresholds: attention, graph min-cut, sparse solvers, tiered compression | [crates.io](https://crates.io/crates/wifi-densepose-ruvector) |
513514
| [Self-Learning WiFi AI (ADR-024)](#self-learning-wifi-ai-adr-024) | Contrastive self-supervised learning, room fingerprinting, anomaly detection, 55 KB model | [ADR-024](docs/adr/ADR-024-contrastive-csi-embedding-model.md) |
515+
| [Cross-Environment Generalization (ADR-027)](#cross-environment-generalization-adr-027) | Domain-adversarial training, geometry-conditioned inference, hardware normalization, zero-shot deployment | [ADR-027](docs/adr/ADR-027-cross-environment-domain-generalization.md) |
514516

515517
</details>
516518

‎docs/adr/ADR-002-ruvector-rvf-integration-strategy.md‎

Lines changed: 3 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,7 +1,9 @@
11
# ADR-002: RuVector RVF Integration Strategy
22

33
## Status
4-
Proposed
4+
Superseded by [ADR-016](ADR-016-ruvector-integration.md) and [ADR-017](ADR-017-ruvector-signal-mat-integration.md)
5+
6+
> **Note:** The vision in this ADR has been fully realized. ADR-016 integrates all 5 RuVector crates into the training pipeline. ADR-017 adds 7 signal + MAT integration points. The `wifi-densepose-ruvector` crate is [published on crates.io](https://crates.io/crates/wifi-densepose-ruvector). See also [ADR-027](ADR-027-cross-environment-domain-generalization.md) for how RuVector is extended with domain generalization.
57
68
## Date
79
2026-02-28

‎docs/adr/ADR-004-hnsw-vector-search-fingerprinting.md‎

Lines changed: 3 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,7 +1,9 @@
11
# ADR-004: HNSW Vector Search for Signal Fingerprinting
22

33
## Status
4-
Proposed
4+
Partially realized by [ADR-024](ADR-024-contrastive-csi-embedding-model.md); extended by [ADR-027](ADR-027-cross-environment-domain-generalization.md)
5+
6+
> **Note:** ADR-024 (AETHER) implements HNSW-compatible fingerprint indices with 4 index types. ADR-027 (MERIDIAN) extends this with domain-disentangled embeddings so fingerprints match across environments, not just within a single room.
57
68
## Date
79
2026-02-28

‎docs/adr/ADR-005-sona-self-learning-pose-estimation.md‎

Lines changed: 3 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,7 +1,9 @@
11
# ADR-005: SONA Self-Learning for Pose Estimation
22

33
## Status
4-
Proposed
4+
Partially realized in [ADR-023](ADR-023-trained-densepose-model-ruvector-pipeline.md); extended by [ADR-027](ADR-027-cross-environment-domain-generalization.md)
5+
6+
> **Note:** ADR-023 implements SONA with MicroLoRA rank-4 adapters and EWC++ memory preservation. ADR-027 (MERIDIAN) extends SONA with unsupervised rapid adaptation: 10 seconds of unlabeled WiFi data in a new room automatically generates environment-specific LoRA weights via contrastive test-time training.
57
68
## Date
79
2026-02-28

‎docs/adr/ADR-006-gnn-enhanced-csi-pattern-recognition.md‎

Lines changed: 3 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,7 +1,9 @@
11
# ADR-006: GNN-Enhanced CSI Pattern Recognition
22

33
## Status
4-
Proposed
4+
Partially realized in [ADR-023](ADR-023-trained-densepose-model-ruvector-pipeline.md); extended by [ADR-027](ADR-027-cross-environment-domain-generalization.md)
5+
6+
> **Note:** ADR-023 implements a 2-layer GCN on the COCO skeleton graph for spatial reasoning. ADR-027 (MERIDIAN) adds domain-adversarial regularization via a gradient reversal layer that forces the GCN to learn environment-invariant graph features, shedding room-specific multipath patterns.
57
68
## Date
79
2026-02-28

0 commit comments

Comments
 (0)