- Fix Viz3D exports to include flow visualization functions
- Export setFlowLayerVisible, setDwellLayerVisible, setCorridorLayerVisible
- Export setFlowTimeFilter, setFlowData, setDwellData, setCorridorData
- Remove duplicate setDwellLayerVisible function definition
This completes the crowd flow visualization feature that was
already implemented in the backend (flow.go) and frontend
(crowdflow.js, viz3d.js) but had missing exports in the Viz3D module.
Implement localization that learns from ground truth data.
- BLE integration as ground truth source: BLE RSSI trilateration provides
continuous position estimates for registered devices
- Fresnel zone weight refinement: SGD-based per-link weight learning
- Continuous weight adjustment based on BLE-blob position feedback
- SelfImprovingLocalizer ties together BLE ground truth, weight learning,
and fusion engine
- REST API endpoints for weights, ground truth, accuracy tracking
Acceptance: Localization accuracy improves automatically as BLE ground
truth data accumulates.
Implement REST API endpoints for managing learned weights and tracking
improvement in the self-improving localization system.
- Add LocalizationHandler with endpoints for:
- GET /api/localization/weights - get all learned link weights
- GET /api/localization/weights/{linkID} - get specific link weight
- POST /api/localization/weights/reset - reset all weights to default
- GET /api/localization/spatial-weights - get spatial weights per zone
- GET /api/localization/groundtruth/* - ground truth sample management
- GET /api/localization/accuracy/* - position accuracy tracking
- GET /api/localization/learning/* - learning progress and history
- Integrate spatial weight learner into fusion engine:
- Add AddLinkInfluenceWithSpatialWeights to grid.go for per-cell weight application
- Update Fuse() in fusion.go to use spatial weight functions when available
- Apply both sigma adjustments and spatial weights for Fresnel zone computation
- Add comprehensive table-driven tests for all API endpoints
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Implement a complete self-improving localization system that uses BLE
RSSI triangulation as ground truth to learn per-link, per-zone weights
for the Fresnel zone fusion engine.
Key components:
- Ground truth sample collection with confidence > 0.7 and distance < 0.5m gates
- Spatial weight learner using online SGD with L2 regularization
- Validation gate that rejects updates without 5% improvement on holdout set
- Bilinear interpolation for smooth weight transitions between zones
- SQLite persistence for weights with 10,000 sample cap per person
- Position accuracy trend visualization in dashboard Accuracy panel
Backend (mothership/internal/localization/):
- groundtruth.go: BLE trilateration provider with Gauss-Newton optimization
- groundtruth_store.go: SQLite storage with weekly accuracy rollups
- spatial_weights.go: SpatialWeightLearner with SGD, validation, interpolation
- weightlearner.go: WeightLearner with error history tracking
- weightstore.go: Weight persistence to SQLite
Frontend (dashboard/js/accuracy.js):
- fetchPositionAccuracy/fetchPositionHistory functions
- drawPositionSparkline for weekly median error trend
- Position accuracy section with median error, trend indicator, sample count
Tests cover:
- Sample collection gates (confidence/distance thresholds)
- SGD weight updates after 100 samples
- Validation gate rejection of adversarial samples
- Bilinear interpolation smoothness
- SQLite sample cap enforcement
- Weight persistence across restarts
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Dashboard hub broadcasts motion state changes immediately on transition
(idle↔motion) via BroadcastMotionState; periodic state snapshots include
motion_states for new client init
- Per-link presence badge (green CLEAR / red MOTION) rendered in link list
alongside global presence indicator in status bar
- Amplitude mean time-series chart (60 s rolling window) for selected link,
line segments colored by motion state at each sample
- Fix: links created from JSON link_active/state events now initialize
ampHistory and lastAmpSample so time-series accumulates from first frame
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
