Tier 1 (Q3 2026): Ship the 4-item plan-antennas / pose-PABS / cog-count v0.0.3 / ADR-029 amendment

What this issue is about (plain-language intro)

This is the Q3 2026 implementation issue spun out of the #748. The research loop ran on 2026-05-22 and produced four high-leverage Tier 1 items that can ship in the next quarter for a total of ~490 LOC, 3-4 person-weeks of engineering.

Each Tier 1 item delivers a measurable user-visible improvement and unblocks the higher-tier work. After this issue is closed, RuView customers will have:

• Sensing coverage that is 93x better than the current "stick the antennas anywhere" recipe
• Intruder / fall detection at 9.36x lift over the previous baseline
• ADR-029 multistatic spec is no longer open-ended on placement

Tier 1 items (in priority order)

1.1 — wifi-densepose plan-antennas CLI tool

Source: R6.2 family (9 sub-ticks), ADR-113 4-axis decision matrix

What it does: Given a room geometry + target zones + cog name, recommend Tx/Rx anchor placement that maximises Fresnel-envelope coverage. Outputs a placement diagram.

LOC: ~360
Effort: ~1 week
Reference code: examples/research-sota/02-placement/ (all 7 scripts)
Architectural decision: docs/adr/ADR-113-multistatic-placement-strategy.md

CLI shape:

wifi-densepose plan-antennas \
  --room 5 5 [Z] \
  --target NAME X Y W H [DX DY DZ] \
  --target-mode {body, chest} \
  --cog COG_NAME \
  --freq-ghz 2.4 \
  --n-anchors N

Expected impact: 93x sensing-coverage lift on the 5x5 m bedroom benchmark. Existing customers can re-mount their seeds today and get 10-100x better sensing without firmware/model changes.

Acceptance criteria:

• CLI runs from repo root in <2 seconds for a 5x5 m room
• Reproduces the headline numbers from R6.2.5 (100% coverage @ N=5 for 4-occupant chest-centric)
• --cog flag auto-selects target-mode + N from ADR-113 matrix
• Output includes placement diagram (ASCII or SVG)
• Documented in user guide

1.2 — R12.1 pose-PABS in vital_signs cog

Source: R12 arc (3 ticks) — R12 NEGATIVE -> R12 PABS POSITIVE -> R12.1 CLOSED LOOP

What it does: Per-frame structural-anomaly detection (intruder, fallen person, new furniture). Couples the pose tracker (ADR-079/101) with R6.1 multi-scatterer forward operator. Subject motion is absorbed into the prediction; only unexplained residuals trigger structure events.

LOC: ~80
Effort: ~3 days
Reference code: examples/research-sota/06-structure-detection/r12_1_pose_pabs_loop.py
Architectural decision: ADR-029 multistatic should reference PABS

Production Rust glue (~80 LOC):

let pose = pose_tracker.estimate(csi_window)?;          // ADR-079 / ADR-101
let expected_scene = body_model.from_pose(pose) + room_walls;
let y_predicted = fresnel_forward.simulate(expected_scene);
let pabs = (csi_window - y_predicted).norm_sq() / csi_window.norm_sq();
if pabs > threshold {
    emit_structure_event();
}

Expected impact: 9.36x intruder detection lift in dynamic scenes (subject walking). False-alarm problem from naive PABS resolved.

Acceptance criteria:

• Slots into existing vital_signs cog per-frame inference path
• Reproduces 9.36x lift on synthetic walking-subject + intruder benchmark
• Configurable threshold + per-frame plausibility window
• Emits structured event when threshold crossed

1.3 — cog-person-count v0.0.3 with chest-centric placement

Source: R5 saliency + R8 RSSI + R6.2.3 chest-centric + ADR-113

What it does: Update existing cog-person-count (ADR-103) to use chest-centric target zones from ADR-113 matrix. Re-train with placement-aware data augmentation.

LOC: ~50 (placement-aware training config + per-cog default --target-mode=body)
Effort: ~3 days (plus retrain time on RTX 5080)
Reference code: examples/research-sota/04-rssi/r8_rssi_only_count.py
Depends on: 1.1 CLI tool (validates placement against matrix)

Expected impact: v0.0.2 + chest-centric placement should improve class-1 accuracy from 34.3% to ~50%+ by removing limb-confound (R6.1 4.7 dB penalty).

Acceptance criteria:

• Trained model checkpoint at v2/crates/cog-person-count/cog/artifacts/count_v0_0_3.safetensors
• Manifest signed per ADR-100 (Ed25519 today; dual-sig per ADR-109 when Phase 1 ships)
• Bench-validated on Cognitum-cluster Pi cluster
• Performance numbers logged for federation roadmap

1.4 — ADR-029 amendment with ADR-113 placement matrix

Source: ADR-113

What it does: Update docs/adr/ADR-029-*.md (RuvSense multistatic) to reference the ADR-113 4-axis decision matrix. Closes the open question that ADR-029 left around anchor placement.

LOC: 0 (ADR-authoring time only)
Effort: ~1 day
Architectural decision: Already in ADR-113; just needs the cross-reference

Acceptance criteria:

• ADR-029 status updated from "Proposed" to "Proposed + amended"
• Cross-reference table to ADR-113 added
• Anchor-count recommendation per cog category included

Critical path

1.1 plan-antennas CLI ────┬──> 1.3 cog-person-count v0.0.3
                          │
1.2 R12.1 vital_signs ────┘
                          
1.4 ADR-029 amendment (independent)

1.1 + 1.2 can run in parallel. 1.3 depends on 1.1 (validation). 1.4 is independent.

What is NOT in this issue

• Tier 2-5 work (federation, PQC, cross-room re-ID, vertical cogs)
• Bench validation (Tier 2.3, separate)
• Regulatory pathway for healthcare cogs (Tier 6.3, separate)
• Real NV-diamond hardware integration (Tier 4+, ADR-114 follow-up)

How to claim a Tier 1 item

1. Comment on this issue with the item number (e.g. "claiming 1.1")
2. Open a feature branch
3. Implement per the acceptance criteria
4. PR-then-auto-merge per the loop's pattern
5. Update the checklist on this issue

Estimated total Tier 1 budget

Item	LOC	Days	Priority
1.1 CLI tool	360	5	HIGH
1.2 R12.1 in vital_signs	80	3	HIGH
1.3 cog-person-count v0.0.3	50	3 (+ retrain time)	HIGH
1.4 ADR-029 amendment	0	1	HIGH
Tier 1 total	~490 LOC	~12 person-days

References

• Entry-point issue: linked above
• Full production roadmap: docs/research/sota-2026-05-22/PRODUCTION-ROADMAP.md
• Loop final summary: docs/research/sota-2026-05-22/00-summary.md
• Architectural decisions: docs/adr/ADR-105 through ADR-114
• Reference numpy implementations: examples/research-sota/

Choose one of the four independently listed Tier 1 items before starting. For antenna planning, read docs/adr/ADR-113-multistatic-placement-strategy.md and examples/research-sota/02-placement/; for pose-PABS, start with examples/research-sota/06-structure-detection/r12_1_pose_pabs_loop.py and the existing vital_signs path. Done means the selected item meets its checklist, including the named benchmark, artifact, event behavior, or ADR documentation outcome.