Updated the meta-rule document to emphasize strict adherence to skill instructions, prohibiting unnecessary investigations or external checks. Revised acceptance criteria and restrictions to correct communication protocol details for ArduPilot and iNav, ensuring clarity on external-positioning interfaces. Adjusted autodev state to reflect ongoing research phase and updated sub-step details for improved tracking.
1.6 KiB
Autodev State
Current Step
flow: greenfield
step: 2
name: Research
status: in_progress
sub_step:
phase: 10
name: sq3-sq4-c1-vio-candidates
detail: "Mode A Phase 2 engine Step 4 — SQ3+SQ4 per-component candidate matrix, starting with C1 (VO/VIO). SQ1 ✓, SQ2 ✓ closed with three architectural decisions resolved by user (recorded in 00_question_decomposition.md): (1) DSM dependency = (a) 3-DoF acceptance (no DSM in cache; attitude from IMU/VIO); (2) AdHoP = (b) conditional (only when initial reprojection error exceeds threshold); (3) Top-N re-rank = (a) promoted to explicit pipeline sub-stage between C2 and C3. Pipeline shape entering SQ3+SQ4: C1 (VIO) → C2 (VPR) → Top-N re-rank → C3 (matcher) → AdHoP-conditional → C4 (PnP+RANSAC+LM) → C5 (estimator) → C8 (FC adapter), with C6 (cache 2D-ortho) + C7 (Jetson runtime) + C9 (datasets) + C10 (provisioning) cross-cutting. C1 candidates entering SQ3+SQ4 (RTAB-Map and ORB-SLAM3 already pruned by Fact #16): VINS-Mono / VINS-Fusion / OpenVINS / OKVIS2 / DROID-SLAM / DPVO / pure-VO baseline (KLT + RANSAC homography). Per-mode context7 capability verification mandatory for every lead library/SDK candidate. Next conversation: re-enter Step 2 / sub_step phase 10 name 'sq3-sq4-c1-vio-candidates', execute Per-Component Search Plan (research/steps/03_engine-investigation.md Step 4) starting with C1: enumerate 5+ search variants per candidate, log to 01_source_registry.md + 02_fact_cards.md, then Component Applicability Gate (must pass: monocular VIO, ≥3 fps output, Jetson Orin Nano runnable, license OK for dual-use)."
retry_count: 0
cycle: 1