OpenSynaptic Cloud — road risk + urban environmental data platform,
powered by mobile bus sensing and edge AI.
道路风险 + 城市环境数据平台 — 公交移动感知 + 边缘 AI。
OpenSynaptic Cloud is the platform layer of the FleetSense ecosystem. Public buses equipped with edge AI modules detect road damage and measure environmental conditions — all processed on-device with privacy-by-design. Raw video never leaves the vehicle.
The platform aggregates anonymized telemetry into two data products:
- Road segment risk index — contract-scoped risk ranking from open-data substitutes + FEM simulation exposure
- Urban heat island layer — block-level air temperature distribution from mobile sensing + machine learning (R²=0.893)
| Layer | Component | Purpose |
|---|---|---|
| Edge | Bus-mounted module (Rock 5T + YOLO) | Road damage detection, T/RH sensing, on-device anonymization |
| Protocol | OpenSynaptic 2-N-2 | UCUM units · Base62 compression · binary packets · CRC-8/16 |
| Transport | UDP / TCP / MQTT / LoRa / BLE | Pluggable transporters for any network |
| Platform | Cloud Station (Rust/Axum) | Identity, RBAC, business proxy, admin panel, data ingestion |
| Business | risk-model (Python) + heat-island (ML) | Risk scoring, FEM simulation, UHI prediction |
| Data | PostgreSQL (AWS RDS) | Persistent storage with TLS |
| Edge AI | YOLOv8 (ONNX) | Road damage classification: cracks, potholes, alligator |
| Product | Customer | Form | Accuracy |
|---|---|---|---|
| Road segment risk index | Municipal, maintenance contractors | REST API + dashboard | L0 evidence level (discussion aid) |
| Urban heat island layer | Urban planners, environmental researchers | Map layer + API + reports | R²=0.893, MAE=0.57°C (ML model) |
| Road condition events | Map / mobility services | REST JSON (real-time) | YOLO detection + human review loop |
| Contract risk reports | Bidding / procurement teams | PDF + API | Open-data substitutes + FEM exposure |
| Component | Technology |
|---|---|
| Edge processing | Rock 5T (RK3588) · YOLOv8 ONNX · OpenSynaptic-TX (C89, ~100B RAM) |
| IoT protocol | OpenSynaptic 2-N-2 · UCUM · Base62 · binary packets · 8-bit MCU compatible |
| Cloud platform | Rust (Axum) · PostgreSQL (RDS) · Cloudflare (DNS + TLS + AOP) |
| Business logic | Python (risk-model) · FEM simulation · XGBoost/RF (UHI model) |
| Web frontend | 3-layer HTML templates · CSP compliant · dark/light theme · responsive |
| Data pipeline | HKO Open Data · HK GeoData Store · synthetic UHI physical model |
- Raw video never leaves the device — only structured events + road-damage crops
- No passenger face or identity processing
- Coarse geolocation (road segment level, not precise track)
- Image crops exclude people, vehicles, license plates
- Auditable event fields — all data ingestion logged
- Invite-gated access — role-based permissions (viewer → admin)
L0 discussion aid · not statutory inspection · not miss-rate · risk ≠ health
The OpenSynaptic protocol stack is open-source, spanning from 8-bit MCUs to desktop dashboards:
| Component | Standard | RAM | Purpose |
|---|---|---|---|
| OSynaptic-TX | C89 | ~100B | Transmit-only encoder (3 API modes) |
| OSynaptic-RX | C89 | ~200B | Receive-only decoder |
| OSynaptic-FX | C99 | ~2KB | Full encode+decode for MCUs |
| OpenSynaptic Python | Python 3.11+ | — | Reference implementation + CLI/TUI/Web |
| Conformance Suite | L1-L5 profiles | — | Wire compatibility certification |
All implementations produce identical binary wire frames — from ATmega328P to cloud servers.
For pilot partnerships, data product inquiries, or technical collaboration:
- Platform: opensynaptic.cloud
- API Documentation: /docs/api
- System Status: /status
- Registration: Invite-gated — contact the team for access