- Jump to see the final results here: Final Results
My thermostat currently a problem: it is outdated, clunky and not connected to the internet. This project's aim is to develop a way to make it wifi controllable, making it easily accessible from anywhere.
To keep things as modular as possible, the plan is to develop an easily attachable mount which connects to my thermostat dial using a simple gear system. A servo will control the positioning of the gears, and since the system is stationed in-place only through the use of double-sided tape, it will be trivial to remove when necessary.
In order to make it a smart machine, we will use Viam's platform along with their TypeScript SDK, allowing for a peer-to-peer connection with the Raspberry Pi.
It's important to be able to test the viability early on, so I started off by prototyping the gears and validating whether the servo is capable of overcoming the inherent stiffness in the dial.
Since the thermostat dial can be difficult to turn at times, a gear ratio of 1.3 provides some extra torque to hopefully ensure that our small servo doesn't struggle.
However, since the servo turns a maximum of 180 degrees, we can only rotate the dial itself about
(Future improvements could have a multi-turn servo or another servo with a wider range that enables selecting any temperature on the thermostat.)
After measuring out the dimensions and doing some 3D modeling work, we have some gears ready to print. These do usually take a few attempts as the plastic shrinks when it cools, but fortunately, only one failed attempt later, we have a perfect fit around the thermostat!
My next step was getting the Raspberry Pi set up with viam-server and attempting to control the servo's angle.
In this case, I called my robot 'thermo' and initialized components for the board itself as well as the servo. Doing so required specifying a maximum range for the servo angle in addition to the gpio pin number it was connected to.
3443.mp4
Above demonstrates testing the servo moving between different angles. Furthermore, I was able to verify the servo's strength by holding it against the dial to test whether the servo would in fact cause it to spin. Although the servo appears small, fortunately it's more than powerful enough.
Encouraged by the servo's movement, it was time to build an enclosure to mount my smart machine, consisting of the Raspberry Pi and servo, alongside the thermostat. This would be the final construction step, and afterward the only remaining piece is the web application.
A few more hours of 3D modeling work later, and we have a simple yet functional enclosure to hold everything together.
The following morning, with the 3D print complete and time ticking down, I placed the components inside the enclosure and mounted it beside my thermostat. The robot was now ready to power on, and hopefully I would never need to manually adjust the dial again!
3454.mp4
Through the Viam web interface, I can debug to adjust the servo angle, however, I wanted to integrate with the SDK, and develop my own site customized to control this thermostat robot.
The website code is all contained in this repository. It was built on top of React, Next.js, and Tailwind
My initial brainstorm for the UI was to create a digital representation of the thermostat dial, reflecting its real-world state. In addition, I'll add some buttons for sending commands to adjust the temperature in my room by adjusting the servo angle.
Upon opening the site, we are greeted with a page asking for your API Key and API Key ID. With the valid credentials provided, we see a virtual representation of the physical thermostat in my room!
The currently set temperature and the dial's current position are visible, all based on the servo angle data fetched through the SDK. As we press the buttons to select a new temperature, we can see the animation as it spins to our desired position. All of this happens instantly, reflecting the actual thermostat's position in my room.
example-dial.mp4
This is made possible by Viam, which creates a low-latency WebRTC connection directly to the Raspberry Pi. The commands are encoded using protobuf then sent through gRPC to call the necessary functions to move the servo at the desired angle.
Here is the final result and a full side-by-side of my actual thermostat being controlled by the web application. The clicks on the UI are nearly instantaneously sent to the servo. No more needing to get out of bed on a cold night to dial up the temperature!
side-by-side.mp4
Another flaw in my thermostat is the inadequate temperature sensor. It is incorrect to within a few degrees, which is enough to make the room uncomfortably hot / cold. Integrating a more accurate sensor with thermo will allow it to autonomously dial the thermostat up when the room needs to be heated and otherwise set the thermostat off.