🚚 FleetVision AI — Smart Fleet Monitoring & Predictive Maintenance

Empowering data-driven fleet operations through intelligent automation, real-time analytics, and predictive maintenance powered by AI.

---

🧩 Problem Statement

Fleet operations across logistics, transportation, and delivery industries face critical challenges:

• Inefficient monitoring of vehicles in real-time.
• Reactive maintenance, resulting in costly downtime.
• Fragmented data systems, preventing comprehensive analysis of fleet health.
• Lack of predictive insights, leading to unexpected failures and financial losses.

FleetVision AI emerges as a solution that integrates telemetry, automation, and AI analytics into a single platform. By capturing sensor data in real time and applying machine learning for anomaly detection and predictive maintenance, FleetVision AI helps organizations reduce operational costs, prevent mechanical failures, and increase safety and efficiency.

---

⚙️ Stack Overview

Stack: FastAPI (REST API) · MySQL 8 · SQLAlchemy · Streamlit (Dashboard) · scikit-learn (IsolationForest) · Docker

Layer	Technology	Description
Backend API	FastAPI + SQLAlchemy	High-performance REST endpoints for data ingestion and model training
Database	MySQL 8	Reliable telemetry storage with structured schema
Machine Learning	scikit-learn	Isolation Forest for anomaly detection and maintenance forecasting
Frontend Dashboard	Streamlit + Plotly + Folium	Interactive web dashboard with real-time KPIs, maps, and alerts
Infrastructure	Docker Compose	Fully containerized environment for reproducibility
Simulation	Python	Synthetic GPS and sensor telemetry generator

---

✅ Core Features

• 📡 Real-time telemetry ingestion via REST API.
• 🧠 AI-driven anomaly detection with Isolation Forest.
• 🗺️ Interactive dashboard: live map, performance KPIs, and time-series analytics.
• ⚙️ On-demand model retraining (/train endpoint).
• 🧰 Telemetry simulator (speed, temperature, fuel, GPS).
• 🐳 Docker Compose deployment (MySQL, API, Dashboard).

---

▶️ Quick Start (Docker)

# 1. Enter the project directory
cd fleetvision

# 2. Build and run all services
docker compose up --build

# 3. Open the dashboard
http://localhost:8501

# 4. (Optional) Run the simulator locally
pip install -r simulator/requirements.txt
python simulator/simulate.py --api http://localhost:8000 --vehicles 3 --interval 1.5

The backend automatically creates the required MySQL tables on startup. Database persistence is ensured through the dbdata Docker volume.

---

🔌 REST API Overview

Method	Endpoint	Description
GET	/health	Check API status
POST	/ingest	Send telemetry data
GET	/vehicles	Retrieve active vehicle IDs
GET	/telemetry?vehicle_id=TRUCK-01&limit=200	Fetch recent telemetry records
POST	/train	Retrain anomaly detection model

Example /ingest Payload

{
  "vehicle_id": "TRUCK-01",
  "timestamp": "2025-10-08T18:00:00Z",
  "lat": 19.4326,
  "lon": -99.1332,
  "speed": 45.2,
  "engine_temp": 88.5,
  "fuel_level": 78.9,
  "accel": 0.12
}

---

🧠 Machine Learning

• Model: Isolation Forest (n_estimators=200, contamination=0.05).
• Features: speed, engine_temp, fuel_level, accel.
• Score normalization: decision_function → [0,1]; scores > 0.6 indicate anomalies.
• Retraining: available through /train endpoint, using stored telemetry data.

---

📊 Future Enhancements

• 🔄 Integration with TimescaleDB for advanced time-series analysis.
• ⚡ Async job handling via Celery + Redis.
• 📲 Real-time alerts through Twilio SMS/WhatsApp or email.
• 🤖 Predictive maintenance using XGBoost and Prophet models.
• 🔐 Authentication (JWT) and multi-organization architecture.

---

🧾 License

This project is licensed under the MIT License.

---

👥 Credits

Developed by Luis Ángel Pérez Castro 📍 Mechatronics Engineer & Fullstack AI Developer (UNAM – Facultad de Ingeniería)

Contributors welcome — whether you’re interested in machine learning, automation, or full-stack development, you can help expand FleetVision AI into a robust platform for intelligent fleet management.