Building Internet Enabled Things with Arduino XBee and Nodejs


I've been building my own internet enabled thermostat to control the heating and cooling systems of my house. This was originally sparked by the Nest thermostat, but I quickly lost interest after discovering the cost. Part of the problem is that I have a split system. My furnace controls are on the first floor and my cooling controls are on the third. At 250 dollars a pop, getting two to cover my whole house seemed a little cost prohibitive. My house is also 3 stories, and most of the time I only care about the temperature of a certain floor. After trying to find something cheaper, I chose to build something on my own using arduino with xbee radios and a raspberrypi. I split up the architecture into sensors and a gateway so I could get the benefits of a zigbee mesh network, and still have the computing power of a linux system on the gateway.

For the time being I'm calling it my home monitoring platform. It's a bad name, but I'm not in marketing. Its built using nodejs, mongodb, and arduino. In the future I'll use redis for scaling out the connections, and for processing events. You can follow the development here:

Home-monitor is a 3 layered project, each layer communicates with the layer above it, and below it.

The first and bottom layer is the sensing layer, this is where sensors live. Sensors emit packets of defined framing and formats for announcing their status. They can also receive packets to do actions. I've only created the thermostat sensor so far. Here is the thermostat: Arduino Xbee Thermostat It's made of an arduino uno, a 2 relay xbee sheild, an xbee, a temperature/humidity sensor, and a few leds. The arduino sketch and the xbee sensor set up are on the repo.

The second layer or application is the sensor-communicator which runs on a raspberrypi with an xbee explorer dongle. I chose the raspberrypi with and xbee explorer because the software runs fine on my laptop too. The sensor-communicator has a few jobs.

  1. Discovers sensor nodes.
  2. Receives and converts sensor readings to JSON.
  3. Performs the OAuth handshake with the cloud-coordinator.
  4. Gets a encrypted key from the cloud-coordinator, using the OAuth protected JSON API.
  5. Connects with and sends the encrypted key over the connection.
  6. Reports sensor readings to the cloud, when they differ from the last reading.
  7. Receives actions over the connection and turns the JSON formatted action to a binary buffer and sends it to the correct sensor.

The sensor-communicator essentially acts as a communication trunk for sensors to the internet. Here is my raspberrypi with the xbee explorer: Raspberrypi + Xbee The final layer is the cloud-coordinator. The sensor-communicators connect to this, and report what sensors they are in charge of, and also authenticate against this. It has a bunch of responsibilities as well:

  1. Act as a OAuth gateway to allow access for sensor-communicators.
  2. Sign up and Log in users.
  3. Serve Web and Mobile UI.
  4. Allow viewing of sensor readings and interaction with sensors.
  5. Expose a JSON api for other services that want to interact with sensors (Protected by OAuth)
  6. Sends actions down to sensor-communicators over connection

I started making the project to deal with my internet thermostat use case and quickly realized that I could extend it to support any type of sensor I could think up. To do that I wanted to create an easy to use JSON format to represent sensors. Here is an example of my thermostat:

  1 : {
    sensorType: 'thermostat',
            type : 'uint8',
            currentTemp : 'float32',
            setTemp: 'float32',
            setting : 'uint8',
            relay1 : 'uint8'
      {name: 'Current Temp', key: 'currentTemp', type:'degree'}
    , {name: 'Set Temperature', key: 'setTemp', type: 'degree'}
    , {name: 'Setting', key : 'setting', type: 'enum',  keys: [0,1,2], values:['auto', 'off', 'on']}
    , {name: 'Relay Power', key : 'relay1', type: 'enum',  keys: [0,1], values:['off', 'on']}
      {  name: 'set',
        type: 0,
          {name: 'state', formtype: 'dropdown', valuetype: 'uint8', settings:['auto', 'off', 'on'], values:[0, 1, 2]},
          {name: 'setTemp', formtype: 'slider', valuetype:'float32', min:50, max:80, value:60 }
        order: ['state', 'setTemp']

the 1 denotes the first byte of a sensor reading, this allows me to look up the rest of the format in my sensorData.js file. (There is only one type right now) The reading object is used to decode the rest of the fields in the packet. I'm using to parse the binary packet to a nice JSON payload. The sensor communicator will then take this JSON payload and PUT it on the sensor-coordinator. The display field is used to display sensor readings. Then a jade template convert fields to html. Eventually I hope to cover a whole bunch of types of fields. After the latest sensor reading has been formated it will append a button for the actions in the action array. I've created a similar jade template for creating forms based off the JSON in action. The action field is also used on the sensor-communicator to translate a JSON payload back to a binary packet that gets sent to the sensor. I hope that when I want to add another sensor the json format will help me quickly add one.

Here's a video of all of that working:

I'm still working on the web and mobile UI's. I've recently started using Backbone.js at work and I plan on using it to make a more polished and scalable UI. I'm also working on a few different types of sensors, my next one will be a Text LCD screen that will report button presses, and display text. Additionally I want to make a rules engines that runs on both the sensor-communicator and cloud-coordinator, This would give me flexibility so I could do rules like:

if(livingroom.temp < 65)

I haven't decided where to put the rules engine, but I think most home automation systems out there live on a split brain type of system.


I added a diagram of what the end architecture will look like. Full Architecture

I also started working on a rules engine. The rules are written in pure JSON, and it can fire off functions which it can reference from a catalog. More on that later.

Leave a comment if you have any suggestions or questions.

Using Upstart and Monit to run nodejs


There are a lot of really great platform as a service offerings

read more

First Post: a simple webrtc chatting service


I've been working a lot with webrtc at work and I wanted to

read more