Smart IoT Temperature and Motion Monitoring System with Machine Learning

Overview

This project is an IoT-based adaptive system that uses temperature and motion sensors to monitor environmental conditions and control LEDs/fans for two different zones (Dog Room and Raz Room). It leverages MicroPython, MQTT for messaging, FastAPI for machine learning-based predictions, and an LCD display to show real-time sensor data. The system supports both manual control (via MQTT commands) and automatic sensor-based control, making it an efficient solution for smart homes or remote monitoring systems.

---

Table of Contents

• Project Features
• System Architecture
• Hardware Setup
• Software Setup
  • Libraries
  • MQTT Broker
  • Machine Learning API
  • Connecting to WiFi
  • Running the Code
• Usage
  • Manual Control via MQTT
  • Data Monitoring
  • Machine Learning Predictions
• Future Improvements
• Contributors

---

Project Features

• Temperature and Humidity Monitoring: Uses DHT22 sensors to measure temperature and humidity for both the Dog Room and Raz Room.
• Motion Detection: PIR motion sensors detect movement in each room, enabling automatic responses.
• MQTT Communication: Real-time control and data publishing using MQTT protocol.
• LED/Fan Control: Adaptive control of LEDs (acting as fans) based on sensor data and user commands.
• Machine Learning Model: Predicts future temperature changes using a FastAPI-based ML model deployed on a remote server.
• LCD Display: Displays current status, sensor readings, and room conditions in real-time.
• Google Home Integration: Supports external control via MQTT master topics for integration with voice assistants.

---

System Architecture

The project architecture involves the following components:

1. Sensors and Actuators (ESP32/ESP8266):

   • Reads temperature, humidity, and motion data.
   • Sends this data to the MQTT broker and LCD.
   • Receives control commands to activate LEDs/fans based on sensor input.

2. MQTT Broker:

   • Handles communication between the ESP devices and other control systems like Google Home or a mobile app.

3. FastAPI-based Machine Learning API:

   • The Server running on a FastAPI, hosts a Random Forest machine learning model that has been trained on Google Colab to predict temperature, humidity and motion based on historical data.
   • The API is queried by the ESP device to adjust the system automatically based on predictions.

---

Hardware Setup

The following hardware components are used in this project:

• ESP32 or ESP8266: For running MicroPython and connecting to the internet.
• DHT22 Sensors (x2): For temperature and humidity measurement in two zones.
• PIR Motion Sensors (x2): To detect movement in the Dog Room and Raz Room.
• LEDs (x2): To represent fans or light control in each room.
• I2C LCD Display (4x20): To show real-time data and status of the system.
• Wires & Breadboard: For connecting sensors, LEDs, and other peripherals.

Pin Configuration:

Component	Pin
DHT Sensor 1
DHT Sensor 2
Red LED (Dog)
Blue LED (Raz)
Motion Sensor 1
Motion Sensor 2
I2C LCD (SCL/SDA)

---

Software Setup

Libraries

The project requires the following libraries to run on MicroPython:

1. umqtt.simple - MQTT client for ESP32.
2. dht - For reading data from DHT22 sensors.
3. led_pwm.py - Custom class for controlling the brightness of the LEDs.
4. i2c_lcd.py - For controlling the I2C LCD display.

MQTT Broker

Broker details in the project:

  MQTT_BROKER = "your_broker_address"
  MQTT_CLIENT = "wokwi001"  # Replace with your device ID

Machine Learning API

The FastAPI-based ML model is deployed on a remote server to predict temperature based on historical sensor data. The ESP32/ESP8266 device sends periodic sensor readings to the API, and in return, the API provides predictions that help the system adjust its behavior (e.g., turning on/off fans, LEDs).

ML API Endpoint:

• URL:
• Method: POST
• Request Format:

  {
    "current_temp": 23.5,
    "current_humidity": 60,
    "room": "dog_room"
  }

• Response Example:

  {
    "predicted_temp": 24.8
  }

Node-Red Dashboard

Connecting to WiFi

Edit the WiFi credentials in the code to connect to your local WiFi:

  WIFI_SSID = "Your-SSID"
  WIFI_PASSWORD = "Your-PASSWORD"

---

Running the Code

1. Flash MicroPython: Ensure your ESP32/ESP8266 board is running MicroPython. You can flash the latest version from MicroPython's official website.
2. Upload the Code: Use tools like Thonny IDE or ampy to upload the .py files to your ESP32/ESP8266 board.
3. Run the Code: After uploading, the device will:
   • Connect to WiFi and the MQTT broker.
   • Start reading data from the DHT22 sensors and motion detectors.
   • Send sensor data to the FastAPI ML server for predictions.
   • Display the data on the LCD and publish it to the MQTT topics.

---

Usage

Manual Control via MQTT

You can manually control the LEDs (fans) via MQTT commands. Send control messages to the iot/device/{device_id}/control topic in JSON format. Examples include:

• Turn on the Red LED (Dog Room):

    {"command": "lamp/red/on"}

• Turn off the Blue LED (Raz Room):

   {"command": "lamp/blue/off"}

• Adjust brightness of LEDs:

   {"command": "lamp/red/brightness/50"}

Data Monitoring

The system continuously sends temperature, humidity, and motion data to the telemetry topic:

{
  "device_id": "wokwi001",
  "temp": 24.5,
  "humidity": 60,
  "motion_dog": "true",
  "motion_raz": "false",
  "type": "sensor"
}

Machine Learning Predictions

The system also queries the FastAPI ML server to receive predictions for real-time temperature adjustments. The predicted data is automatically applied to adjust the LEDs/fans accordingly. You can monitor these predictions via the API logs or a connected MQTT dashboard.

---

Future Improvements

1. Refine the ML Model: Continuously improve the machine learning model with more data for better accuracy in predictions.
2. Add More Sensors: Extend the system to monitor more rooms or environmental factors (CO2, light levels, etc.).
3. Mobile App Interface: Develop a mobile app for better control and visualization of sensor data.

---

Feel free to reach out for feedback or suggestions through my email address at [email protected]