Triton Project is an affordable DIY AR headset

Hello Graham, introduce yourself to my readers!

Hello, my name is Graham Atlee, and I’m a self taught developer. For the past 3 years I’ve been working on various AR/VR projects like Pumori.io and most recently The Triton Project. I like to bootstrap my own products that help inspire people to learn and build from. 

Can you tell us what is the Triton Project?

The Triton Project is a 3D printable AR headset with the ability to assemble right at home using off the shelf parts for a relatively low cost. One of the biggest perks with the Triton is the wide FOV. So when you build a Triton and put it on for the first time, it will feel as if you’re gazing through a pair of AR glasses straight from the future. 

How have you come up with this idea of making an AR headset of your own?

The Triton was inspired by another AR headset known as the Project North Star. I modestly felt as if I could construct a more approachable and miniaturized AR platform. So for 2 weeks working out of my college dorm, I stayed up late almost every night prototyping what became The Triton Project from learning CAD and 3D printing to engineering every angle to fit flawlessly together. Then for the next 6 months I built the platform and documented everything to be replicated by the community both in the user experience and pure excitement as AR brings your surface to life. 

What is the final price if someone wants to build this headset? And what skills are required to assemble it?

The final price averages around $300, but this price is for those that print the 3D parts on their own. If you order from a 3D printing service the price will vary depending on the place. I wouldn’t say it requires heavy technical skills to assemble a Triton, but I do believe that you will see greater success within the project if you have or acquire some background knowledge of 3D printing and CAD (ex. know what a .STL file is, what PLA+ plastic is). These can easily be covered by watching some Youtube videos on it and a little bit of experimenting.

Suppose that I want to build a Triton headset. What are the steps that I have to perform to have one?

First you should go to https://atlee19.github.io/TritonProject-Guide/ 

This is the official documented guide for building a Triton headset. To the right side of the page you will see Build Guides and under it links to assembly, 3d printing, and software setup.

I would start by looking at the bill of materials for the Triton Project and begin ordering everything that is needed before anything else. Some of the parts that require you to order through Aliexpress can take upward of  2+ weeks to be delivered so make sure you factor that into your timeline for the project. TIP: If you can allocate the budget I would recommend buying 2 display LCD panels and 1 driver board. The lcd displays are fragile and are prone to breaking but they are worth it.

Additionally, I would highly recommend buying this head padding kit from Knox Labs as it will dramatically increase comfort as you wear the headset, especially for longer periods of time as you will be immersed within the Triton. 

I do not sell a bundled version of the Triton. You will have to buy the pieces separately and assemble it. I do sell a software launcher that allows you, once the Triton is assembled, to run and begin the software side of the Triton journey. 

Triton works with Leap Motion (now Ultra Leap) hands tracking. Can you tell us about the UX experiments that you have done for AR?

My original concept for The Triton was to create a contextual AR menu from which apps could be loaded.

With Pumori.io, I had created 6 Unity apps that demo UI/UX concepts on the Project North Star headset. However, I had to manually switch between unity packages to demo different apps which led me to taking on and off the headset constantly. 

I wanted to switch apps while inside the headset so I came up with the UX concept above as a new pathway to change quicker and more effective. 

I call this a palm based menu as the whole UI is tracked and displayed onto the palm of your hand. AR apps can be loaded into the 4 cells and activated via clicking on the cells. To close an AR app you squeeze the ball located in the middle of your palm. This leads to less time switching from app to app along with more accuracy and less movement of your hands.

Now 4 app cells may not be an abundance,  but it could theoretically be expanded. However right now the Triton palm menu is just a mere experiment and by no means should be considered a stable UX. 

Do you think you can get rid of the PC connection and make it work as a standalone or connected-to-a-phone device?

I developed this concept diagram up above over a year ago. Originally I was going to make a standalone device which hooked everything up to a Nvidia Jetson Nano that could be worn on your belt (think Magic Leap One). 

The issue fell where I didn’t have a Linux driver for the Leap Motion sensor and there were not many stable alternatives for hand tracking at the time. It also came down to whether the Jetson even had enough power for what I was envisioning. I consulted with Noah Zerkin about it and he said it wouldn’t which confirmed my thoughts. That then put a stop on the standalone concept and I decided to go with a tethered PC experience for the time being. 

What do you think are the main strength point of Triton? And what are its weaknesses instead?

The optical architecture of The Triton is both a huge strength and advantage. The optical reflector found on the Triton is technically called a single large curved-visor combiner, or  “bug eye” combiners for short. The advantage of these combiners is that they yield a decent see-through with reduced distortion, good color fidelity, and a low LCA (lateral chromatic aberration). Additionally the cost of these combiners are quite low ($35-99) which kept the project budget down.

The best feature is the wide FOV which by estimate is about 90° Diagonal. When I put the Triton on, I have the ability to view both hands overlaid with the AR graphics. This is a huge advantage since it allows you to move your hands around freely without the constant worry of staying “in bounds”. 

An AR headset is only as strong as its biggest weakness. Using a single large combiner piece for the imaging task leads to complications. For instance, the display panel had to be angled above the user’s forehead in order for the image to reflect. This causes the center of gravity to then become shifted to the front. Adding more than 1 vision sensor is impossible or else the headset becomes far too front heavy which is a design weakness of The Triton. 

What are the differences and the analogies with the Project North Star?

The North Star is like the iPad and the Triton is the iPad mini. The main structural difference is that the Triton uses a single panel display that is angled above the wearer’s eye line. FOV, resolution, and comfort are all relatively the same.

In terms of software, the North Star uses Unity to render its AR graphics while the Triton uses Three.js. 

I’ve read that the Launcher is in Three.js . Is this the only supported programming language or devs can also use Unity and Unreal Engine?

Yes, Three.js is the only supported library for Triton. At the time, I was trying to stay away from using Unity and .NET because I wanted to keep the development lightweight. This led to me splitting for a web based alternative which was Three.js. The entire stack was written in Javascript which is great of course because everything runs super fast and integrating different APIs was very smooth flowing.

However looking back on it, I think this was a mistake in some ways. The amount of Three.js developers compared to the amount of Unity/Unreal devs is far less. I also realized the development on XR plugins for Unity had advanced and was unifying development for VR. 

I have no plans to provide Unity/Unreal support but if developers want to they could easily make a Unity rig for the Triton.

I’ve read about your long journey in building it. What are the best lessons that you’ve learned while building Triton that you want to share with the community?

There’s actually a blog post that you published in 2018 about the importance of ergonomics in virtual reality that I really agree with. When you go off to construct your own AR/VR headset, especially for the first time, you tend to think in the standard hardware spectrum of things including display resolution, processor speed, etc. 

While those are important factors, there’s a whole other class of engineering problems centered around ergonomics that could make or break an AR/VR headset if overlooked. You really need to be an obsessive ergonomist constantly checking craniofacial measurements, interpupillary distance adjustment, center of gravity, etc.

The mistake I made was thinking that I’m a human therefore designing this headset for my specific human self will solve all things with wearability. That however is not the case across the spectrum. Luckily, I was catering to such a small-scale market where I didn’t have to design for a broader appeal. So if anyone in the community is thinking about designing a headset, you should factorize ergonomics first and fight for as much weight reduction as possible. 

Do you plan starting a Kickstarter campaign for Triton?

No, I don’t plan on starting a kickstarter for the Triton. My plan for The Triton was always to bootstrap it and I have done just that. I wish to take what I’ve learned and then apply it to the next AR/VR product that I work on.

What are the medium and long term goals of Triton?

I think my ambitions for The Triton Project have ultimately peaked. I just want to be a project maintainer or shepherd that guides people in how to build Triton’s. One thing I would love to see more of is the Triton being built in university labs or used in hackathons. 

How can the community help you for this project?

If someone is up for the challenge I would like to see a hacker in the community integrate 3DoF head tracking onto the device. Or even take it 1 step further and use the Intel Realsense to get 6DoF slam tracking working. 

What do you think AR needs to become mainstream?

I think there are still fundamentally unsolved problems in display technology that inhibit the mainstream AR glasses that we all want and envision. The optical architecture I chose for the Triton is obviously very far off from achieving any mainstream usage. However my goal was to give people an impressionable experience that would enable you to project past the current system’s flaws in order to see our AR future.