IoT Smart Room Control System

Team Members:

• Namith Gangireddyvari
• John Peng

Video Demo: Watch on YouTube

---

Project Overview

The IoT Smart Room Control System is an automated environment management solution that regulates temperature and lighting using a distributed "Master-Slave" architecture. The system integrates an Arduino for low-level hardware control and a Raspberry Pi 4 for high-level logic, computer vision, and web hosting.

Users can interact with the system via a Web Dashboard, a physical IR Remote, or enable Auto Mode for fully autonomous operation based on human presence and temperature thresholds.

---

System Architecture

Hardware Stack

• Controller: Raspberry Pi 4 (Control and Logic)
• Microcontroller: Arduino Uno (Outputs/Sensors Interfacing)
• Sensors:
  • DHT11 Temperature & Humidity Sensor
  • Infrared (IR) Receiver
  • USB Webcam
• Actuators:
  • DC Fan Motor (driven by PN2222 Transistor)
  • White LED
  • 9V Battery (Fan Power)

Software Stack

• Backend: Python 3 (Flask, OpenCV, Threading)
• Frontend: HTML5, JavaScript, Chart.js
• Firmware: C++ (Arduino)
• AI Model: YOLOv3-Tiny (Object Detection)

---

Key Features

1. Web Dashboard:
   • Real-time visualization of Temperature and Humidity using Chart.js.
   • Manual toggles for Fan and Light.
   • Adjustable temperature threshold for automatic fan activation.
2. Computer Vision Automation:
   • Uses YOLOv3-Tiny to detect human presence.
   • Auto-Light: Lights turn ON when a person enters and OFF when the room is vacant.
3. Smart Fan Control:
   • Automatically activates the fan when the room temperature exceeds the user-defined threshold.
   • Includes hysteresis logic to prevent motor chattering.
4. Multi-Interface Synchronization:
   • The system utilizes a continuous feedback loop. Whether you use the Web UI or the IR Remote, the system state remains synchronized across all devices.

---

Technical Implementation

Communication Protocol (UART)

The Raspberry Pi and Arduino communicate via serial RX-TX (/dev/serial0) at 9600 baud.

• Arduino -> Pi: Broadcasts JSON status packets every 2 seconds.
  • Example: {"temp":24.5, "humidity":60.2, "fan":true, "led":false}
• Pi -> Arduino: Sends compact string commands for actuation.
  • Example: "FAN:1" or "LED:0"

Multithreading & Concurrency

The Python application handles blocking I/O and heavy processing by running four concurrent threads:

1. Flask API: Handles HTTP requests from the frontend.
2. YOLO Inference: Processes camera frames for human detection.
3. Serial Manager: Listens for Arduino updates and sends commands.
4. Auto Logic: Executes control decisions based on state.

Mutex Locking: state_lock and serial_lock are implemented to prevent race conditions when accessing the shared global state or the serial port.

Optimization

• Buffer Flushing: The vision loop actively discards buffered camera frames to eliminate input lag and ensure real-time responsiveness.
• Resolution: Input frames are resized to 320x320 for optimal performance on the Pi.

---

Installation & Setup

1. Arduino Setup

1. Install the required libraries in the Arduino IDE:
   • DHT sensor library by Adafruit (v1.4.6)
   • IRremote by shiriff, z3t0, ArminJo (v4.5.0)
2. Connect the hardware components according to the circuit diagram.
3. Important: Disconnect the RX/TX pins while uploading the code to avoid interference.
4. Compile and upload the firmware to the Arduino Uno.

2. Raspberry Pi Setup

1. Ensure Python 3 is installed.
2. Install the required Python dependencies:

   pip3 install numpy pyserial flask flask-cors opencv-python

3. Download the required YOLO model files into the project directory:

   curl -L -o yolov3-tiny.cfg "[https://github.com/pjreddie/darknet/raw/master/cfg/yolov3-tiny.cfg](https://github.com/pjreddie/darknet/raw/master/cfg/yolov3-tiny.cfg)"
   curl -L -o yolov3-tiny.weights "[https://pjreddie.com/media/files/yolov3-tiny.weights](https://pjreddie.com/media/files/yolov3-tiny.weights)"
   curl -L -o coco.names "[https://raw.githubusercontent.com/pjreddie/darknet/master/data/coco.names](https://raw.githubusercontent.com/pjreddie/darknet/master/data/coco.names)"

3. Running the System

1. Connect the Arduino to the Raspberry Pi via USB.
2. SSH into the Raspberry Pi.
3. Navigate to the server directory and run:

   python3 full_server.py

4. On a computer connected to the same network, open the frontend.html file in a web browser.

---

Challenges & Limitations

• Electrical Noise: The 9V fan battery shares a common ground with the IR sensor. When the fan is running, it generates electrical noise that the IR sensor interprets as dummy values. A software filter was implemented to ignore these specific noise patterns.
• Blocking I/O: The Flask server is blocking by default. We successfully implemented threading to ensure the camera feed doesn't freeze when the web server is waiting for a request.
• Hardware Uploads: The active serial connection between Pi and Arduino interferes with code uploading; wires must be physically disconnected during firmware updates.

---

External Libraries Used

• Python: threading, time, json, numpy, pyserial, flask, flask-cors, opencv-python
• Frontend: Chart.js
• Arduino: Adafruit DHT, IRremote