Our IoT solution allows users to be in control of air quality in their environments. Users can monitor different parameters about the indoor air quality in real time, or they can view past and (predicted) future values as well by using our web interface. These parameters consist of temperature, humidity, and overall air quality score.

By using machine learning, besides monitoring, our system in the cloud can predict when air quality will drop below optimal levels so that it can send feedback to IoT device by combining real time data and prediction.

• Mert Aközcan
• Emre Bilgili
• Muhammed Emin Güre

• Cloud: Node-RED flow in JSON format (can be imported to any Node-RED flow)
• Figures
  • cloud: Screenshots from Node-RED flow and IBM Bluemix platform
  • configuration_ui: Screenshots of WiFi configuration UI which could be seen when setting up the device for the first time
  • hardware: Pictures of the device and connection diagram
  • web_ui: Screenshots from web UI which includes charts and graphs
• ML: Code for generating test data
• Node: ESP8266 NodeMCU Code written on Arduino IDE
• UI: Code for user interface

• ESP8266 NodeMCU CP2102 V2
• MQ135 Gas Sensor
• DHT22 Temperature and Humidity Sensor
• Buzzer
• Red and Yellow LEDs
• Jumper Cables

• Reading sensor values and publishing them to cloud

loop function in classroom_air.ino reads temperature, humidity and air quality values from temperature&humidity(DHTPIN) and air quality(GASPIN) sensors. Then it prepares a JSON string and publishes that to evt topic in MQTT broker.

• Cloud and data processing

In the Event Subscriber -> Add Timestamp -> Write Database section, timestamp is added to the JSON file. Then, it is written to database.

In the Event Subscriber -> Real Time Analysis -> Quality Publisher section, last gathered data is processed and air quality value is loaded into msg.payload. Here, prediction value is gathered from [get]/api/prediction section and used in the function. Then, air quality value is published to fdb topic in MQTT broker.

• Reading air quality status from cloud and taking actions

receivedCallback function in classroom_air.ino reads air quality status from fdb topic in MQTT broker. If status value is 2 in payload, it makes a "beep beep" sound with buzzer (BUZZERPIN) and lights up the red LED (REDPIN). Else if status value is 1, it lights up the yellow LED (YELLOWPIN).

Returns air quality data within given date interval.

• URL

  api/data

• Method:

  GET

• URL Params

  Required:

  start_date = yyyy-mm-ddThh:mm Beginning of the interval

  end_date = yyyy-mm-ddThh:mm End of the interval

  Optional:

  resolution = integer Average every resolution amount of records into one record

• Success Response:

  • Code: 200
    Content:

    [  
        {  
           "temp":26.120000300000005,
           "hum":57.5700006,
           "quality":425.2,
           "timestamp":1526294065
        },
        {  
           "temp":25.96,
           "hum":58.5099998,
           "quality":424.1,
           "timestamp":1526294165
        },
        {  
           "temp":26.0000002,
           "hum":57.79000020000001,
           "quality":423.7,
           "timestamp":1526294265
        }
    ]
    

• Sample Call:

  curl -XGET 'https://classroom-air-quality-system.eu-gb.mybluemix.net/api/data?start_date=2018-05-14T13:00&end_date=2018-05-14T14:00&resolution=10'

Returns air quality data of last specified hours.

• URL

  api/data

• Method:

  GET

• URL Params

  Required:

  hour = double Number of hours before current time

  Optional:

  resolution = integer Average every resolution amount of records into one record

• Success Response:

  • Code: 200
    Content:

    [  
       {  
          "_id":"f47b3efd8112e46b0764ae7d947ff0d4",
          "_rev":"1-d77399730085967e22a248598f20e792",
          "temp":25,
          "hum":70.699997,
          "quality":240,
          "timestamp":1527353136
       },
       {  
          "_id":"7e91b018cd0a3a25655e81753fc65360",
          "_rev":"1-c85e4c81bb93e90dddca6dd857aeafd9",
          "temp":25,
          "hum":70,
          "quality":345,
          "timestamp":1527353146
       }
    ]
    

• Sample Call:

  curl -XGET 'https://classroom-air-quality-system.eu-gb.mybluemix.net/api/data?hour=1&resolution=5'

Returns the latest air quality data.

• URL

  api/data/latest

• Method:

  GET

• Success Response:

  • Code: 200
    Content:

    [  
       {  
          "_id":"b3d2731c4562041aa3be27e3fbcff3af",
          "_rev":"1-6b92b9f45bf4167c3af14d76de7c5705",
          "temp":25.299999,
          "hum":65.599998,
          "quality":399,
          "timestamp":1527354126
       }
    ]
    

• Sample Call:

  curl -XGET 'https://classroom-air-quality-system.eu-gb.mybluemix.net/api/data/latest'

Returns predicted air quality data for the given day of week within the given time interval. It returns four predictions for each hour in the interval.

• URL

  api/prediction

• Method:

  GET

• URL Params

  Required:

  day = integer Day of the week (Monday: 1, Sunday: 7)

  start_hour = integer Beginning of the interval [0-24]

  end_hour = integer End of the interval [0-24]

• Success Response:

  • Code: 200
    Content:

    [  
       {  
          "quality":429.2912570045413,
          "timestamp":122400
       },
       {  
          "quality":429.28710302441624,
          "timestamp":123300
       },
       {  
          "quality":429.2829490442912,
          "timestamp":124200
       },
       {  
          "quality":429.27879506416605,
          "timestamp":125100
       }
    ]
    

• Sample Call:

  curl -XGET 'https://classroom-air-quality-system.eu-gb.mybluemix.net/api/prediction?day=2&start_hour=10&end_hour=11'

• Operating System: MacOS Sierra 10.13.4
• Tools
  • Arduino IDE 1.8.3
  • IBM Bluemix Internet of Things Platform
  • Node-RED v0.18.4
  • Cloudant NoSQL DB Build 6909
  • IBM Watson Studio (for Machine Learning services)

To setup Arduino IDE in order to compile and upload the code into ESP8266 NodeMCU device,

• Add 'http://arduino.esp8266.com/stable/package_esp8266com_index.json' to Preferences -> Additional Boards Manager URLs
• From Sketch -> Include Library -> Manage Libraries find and install
  • Adafruit Circuit Playground by Adafruit v1.2.1
  • Adafruit Unified Sensor by Adafruit v1.0.2
  • DHT Sensor Library by Adafruit v1.3.0
  • PubSubClient by Nick O'Leary v2.6.0
  • WiFiManager by tzapu v0.12.0

Fake data generator FakeData.py could be run using Python 3.6.4

You may want to modify our project and configure it for your needs. There will be some necessary steps for that.

If you have components listed here,

1. Connect components as described here
2. Download and set up Arduino IDE following these steps
3. Connect your NodeMCU to the computer and upload the code. You may want to change MQTT topics in the code before uploading.
4. After uploading the code, you should connect to WiFi (from your phone or computer) whose SSID is "KURULUM". It will redirect you to select which WiFi AP you want to connect from NodeMCU.
5. To check whether your device is working properly, you can download MQTTBox and subscribe to topic you set up at Step 3. Note that we're using a public broker for MQTT connections.

After completing device set up, you can start working on cloud. We used IBM Blumix IoT Starter Pack as our cloud service.

1. Create an account on IBM Bluemix.
2. Create IoT Starter Pack project and Cloudant NoSQL database.
3. Go to your Node-RED flow and import this code.
4. Change MQTT topics to those you set up when configuring the device.
5. Select your own Cloudant database instance.
6. Modify REST API endpoints as you wish and deploy.

There are many different ways to work with machine learning services so it's better to follow IBM's documentation on this.

To make predictions about air quality level, we used linear regression. Since our case for this project was classroom environment, collected data can be evaluated in weekly basis. We used time passed from the beginning of the week as feature and tried to predict overall air quality score based on the feature.

https://github.com/knolleary/pubsubclient

We used PubSubClient for MQTT connection on ESP8266 NodeMCU device.

https://github.com/tzapu/WiFiManager

With WiFiManager, we can set up WiFi connection of ESP8266 dynamically instead of hard coding that information into the code.

https://www.chartjs.org/

We used Chart.js in our web interface to create beautiful charts and graphics to show air quality data.

https://github.com/puikinsh/Adminator-admin-dashboard

We used Adminator HTML5 template in our web interface to speed up the process with a magnificent UI. The author of this template is Colorlib.