🚗 ADAS Embedded System Graduation Project

A complete Advanced Driver‑Assistance System (ADAS) embedded solution, combining microcontroller firmware, Linux‑based AI processing, over‑the‑air updates, and a unified KivyMD GUI dashboard. Developed at Alexandria University’s Faculty of Engineering, Electronics and Communication Department.

---

📌 Table of Contents

• 🚗 ADAS Embedded System Graduation Project
  • 📌 Table of Contents
  • 📖 Project Idea
  • 🔌 Circuit & Interfacing
  • 🏗 Architecture & Modules
  • 🧠 Key Features
    • ✅ Collision Avoidance
    • ✅ Blind Spot Detection
    • ✅ Drowsiness Detection
    • ✅ Auto Parking (New!)
  • ⚙️ Technologies & Tools
  • 🚀 Getting Started
    • Prerequisites

---

📖 Project Idea

The core idea of this graduation project is to develop a modular, low‑cost ADAS platform that integrates into standard vehicles. The system comprises:

1. AI‑Powered Vision for lane keeping and traffic‑sign recognition on Raspberry Pi 4.

2. Driver Drowsiness Simulation using OpenCV & dlib to emulate in‑cabin monitoring.

3. Embedded Microcontrollers (STM32F103C6) running FreeRTOS for real-time control:

   • Collision Avoidance
   • Blind‑Spot Detection
   • Auto Parking

4. Wireless Bridges using ESP32s for data relaying between modules.

5. GUI Dashboard built with KivyMD on Raspberry Pi 3 for visual feedback.

6. Firmware‑Over‑the‑Air (FOTA) update system hosted locally in the car.

---

🔌 Circuit & Interfacing

[Camera]
   |
   | USB
   v
[Raspberry Pi 4] ── Wi‑Fi MQTT ──► [Raspberry Pi 3 (GUI)]
   │
   │ SPI → ESP32‑ALERT ── Wi‑Fi MQTT ──► [Raspberry Pi 3 (GUI)]
   │
[ESP32‑CTL] ← UART → STM32F103C6 ← SPI → ESP32‑ALERT
   │
[Raspberry Pi 3 (FOTA)] ← UART → STM32F103C6

1. Camera → Raspberry Pi 4

   • USB camera streams video for AI vision (lane/sign) on Pi 4.

2. Raspberry Pi 4 → MQTT → Raspberry Pi 3 (GUI)

   • AI inference alerts (lane, sign, drowsiness) are published directly over Wi‑Fi MQTT.

3. ESP32‑CTL ↔ STM32 (UART)

   • UART at 115200 baud relays mobile‑app control commands to STM32.

4. ESP32‑ALERT ↔ STM32 (SPI)

   • SPI bus gathers collision/blind‑spot data from STM32.
   • Publishes sensor alerts via Wi‑Fi MQTT to the GUI Pi 3.

5. Sensors on STM32

   • 2 × HC‑SR04 ultrasonic sensors (front & rear)
   • 2 × digital IR sensors for side blind‑spot detection

6. Power & Level‑Shifting

   • 10 V → 5 V regulator for all boards.
   • Common ground; level‑shifters ensure safe 3.3 V SPI/UART signalling.

7. Raspberry Pi 3 (FOTA)

   • Hosts Python HTTP server for uploading STM32 .bin files from a laptop.
   • Uses UART to flash new firmware onto STM32.

8. Raspberry Pi 3 (GUI)

   • Runs KivyMD dashboard on a 7″ display.
   • Subscribes to MQTT to show real‑time sensor and vision warnings.

9. Sensor Layout:

   • 4× Ultrasonic sensors for obstacle detection
   • 2× IR proximity sensors for rear corner blind‑spot detection

---

🏗 Architecture & Modules

┌─────────────────────────────────────────────────────────┐
│                     Raspberry Pi 4                      │
│  ┌──────────┐    ┌────────────────────────────────────┐ │
│  │   Yocto  │ →  │     AI_Features (Docker)           │ │
│  │  Linux   │    │    (lane, sign inference)          │ │
│  └──────────┘    └────────────────────────────────────┘ │
│           │ Wi‑Fi MQTT → Raspberry Pi 3 (GUI)           │
│           │                                        ▲    │
│           │ SPI → ESP32‑ALERT ── Wi‑Fi MQTT ── ─ ─ ─ ─┘ │
└─────────────────────────────────────────────────────────┘
           ↓
┌─────────────────────────────────────────────────────────┐
│                  STM32F103C6 Node                       │
│  ┌──────────────┐   ┌─────────────────────────────────┐ │
│  │  Sensors     │   │  FreeRTOS tasks: collision,     │ │
│  │ HC‑SR04 & IR │→  │  blind‑spot, autoparking logic  │ │
│  └──────────────┘   └─────────────────────────────────┘ │
│           ↑ UART → ESP32‑CTL                            │
└─────────────────────────────────────────────────────────┘
           ↓                           
┌────────────────────────┐    
│ Raspberry Pi 3 (GUI)   │    
│ — KivyMD MQTT client   │    
│ — 7″ inch screen       │    
└────────────────────────┘    
           ↓                           
┌──────────────────────────┐    
│ Raspberry Pi 3 (FOTA)    │    
│ — Python HTTP server     │    
│ — UART to STM32 flashing │   
└──────────────────────────┘    

---

🧠 Key Features

✅ Collision Avoidance

Real-time obstacle detection using front and rear ultrasonic sensors. Alerts are sent to the GUI if a collision risk is detected.

✅ Blind Spot Detection

IR sensors monitor the vehicle’s rear corners to detect nearby obstacles in adjacent lanes.

✅ Drowsiness Detection

Computer vision–based eye tracking detects signs of driver fatigue. If detected, an auto-parking request is triggered.

✅ Auto Parking (New!)

The system includes an automatic parking feature that is activated when the drowsiness detection system flags the driver as inattentive. Key aspects:

• Trigger Mechanism: Initiated by an EXTI command from the embedded Linux module upon drowsiness detection.

• Task Coordination: FreeRTOS suspends unrelated tasks (e.g., blind spot & collision detection), then activates:

  • AutoParkingHandler
  • ParkCar

• Operation:

  • Searches for a valid parking space using side-mounted ultrasonic sensors.
  • Executes a reverse-parking maneuver using a sequence of directional motor commands.

---

⚙️ Technologies & Tools

• STM32F103C6 (FreeRTOS-based real-time node)
• Raspberry Pi 3 & 4 (Yocto Linux, Python GUI, Web Server)
• ESP32 (SPI/UART bridge, Wi-Fi MQTT client)
• KivyMD (Python GUI)
• OpenCV + dlib (Drowsiness Detection)
• Docker (for AI Modles deployment)
• Yocto (Linux Image Customization)
• MQTT (HiveMQ) (Pub/Sub communication layer)

---

🚀 Getting Started

Prerequisites

• STM32CubeMX + STM32CubeIDE for STM32 firmware
• Python 3.x with KivyMD for GUI
• Docker for Models Deployment
• Yocto for Linux Customization
• QT Creator for Dashboard Development
• ESP-IDF or Arduino for ESP32 firmware

---