Porter is an autonomous luggage-carrying robot for airports. It navigates terminals using LIDAR, carries passenger luggage, and provides an on-device AI assistant — all running locally on a Raspberry Pi 5. Zero cloud. Zero latency. Full autonomy. Every computation — SLAM, navigation, object detection, natural language — happens at the edge. No internet required. No data leaves the device. | |
┌────────────────────────────────────────┐
│ RASPBERRY PI 5 (Master) │
│ │
┌──────────┐ USB Serial │ ┌──────────┐ /scan ┌──────────┐ │
│ YDLIDAR ├────────────────┼──┤ LIDAR ├──────────►│ Nav2 │ │
│ X4 Pro │ 128000 baud │ │ Driver │ │ SLAM │ │
└──────────┘ │ └──────────┘ └────┬─────┘ │
│ │ │
│ /cmd_vel │ │
│ ┌──────────┐◄───────────────┘ │
│ │ Motor │ │
│ │ Bridge │ ┌──────────────┐ │
│ └────┬─────┘ │ Virtue AI │ │
│ │ │ Qwen 2.5 1.5B│ │
│ │ USB CDC │ 14 Tools │ │
│ │ │ RAG + LoRA │ │
│ │ └──────┬───────┘ │
│ │ │ │
└───────┼──────────────────┼─────────────┘
│ │
┌───────▼────┐ ┌───────▼────────┐
│ ESP32 #1 │ │ Flutter GUI │
│ BTS7960 │ │ Touchscreen │
│ Diff Drive│ │ SSE Streaming │
└────────────┘ └────────────────┘
The robot itself. ROS 2 packages, ESP32 Zephyr firmware, AI assistant, Flutter GUI. Every push builds Docker images, ESP32 Tech: C++17 / Python / Dart / C (Zephyr) | 8 VS Code extensions for the full robotics dev lifecycle — firmware builder, AI studio, ROS 2 node graph, hardware dashboard, simulation manager, PCB studio, and more. Tech: TypeScript / React 18 / React Flow / esbuild |
Production-grade REST API for fleet management with auto-generated Python and Dart client libraries. FastAPI + Pydantic + SSE streaming. Tech: Python (FastAPI) / Dart | Rust-based fleet management backend. Real-time robot telemetry, task assignment, multi-robot coordination. Tech: Rust (axum + tokio + sqlx) |
Deployment profiles, parameter validation, and environment-aware configuration management for all VirtusCo services. Tech: YAML / Python | Architecture docs, API references, development guides, and the repo split plan. Central knowledge base for all VirtusCo projects. Tech: Markdown |
| Layer | Technology | Purpose |
|---|---|---|
| Navigation | ROS 2 Jazzy + Nav2 | SLAM, path planning, obstacle avoidance |
| Perception | YDLIDAR X4 Pro | 360° LIDAR, 12m range, 5000 pts/s |
| AI | Qwen 2.5 1.5B GGUF | On-device NLU, 14 tools, RAG, <1.8s P95 |
| Motor Control | Zephyr RTOS + SMF | Differential drive, PWM, watchdog |
| Sensor Fusion | Zephyr RTOS + Kalman | ToF + Ultrasonic + Microwave fusion |
| GUI | Flutter (Linux) | Passenger touchscreen interface |
| Protocol | CRC16-CCITT Binary | RPi ↔ ESP32 serial communication |
| Fleet | Rust (axum) | Multi-robot coordination backend |
| DevTools | 8 VS Code Extensions | Full robotics development lifecycle |
AI LatencyP50: 1.4sP95: 1.8s | LIDAR360° @ 5kHz12m range | Motor Loop100Hz control50ms watchdog | Sensor Fusion50Hz Kalman3-source |
1. EDGE-FIRST Everything runs locally. No cloud. No latency. No data leaves the device.
2. REAL-TIME Motor control at 100Hz. Sensor fusion at 50Hz. Watchdogs everywhere.
3. FAIL-SAFE 9-state FSM with graceful degradation. Hardware watchdogs. CRC validation.
4. MODULAR Each component is a ROS 2 node. Swap, upgrade, or disable independently.
5. PRODUCTION-GRADE 178 Ztest cases. GTest suites. CI on every push. Automated releases.
Built with purpose. Engineered for reliability. Deployed at the edge.