Supercon 2022: Aedan Cullen Is Creating An AR System To Beat The Big Boys

There’s something very tantalizing about an augmented reality (AR) overlay that can provide information in daily life without having to glance at a smartphone display, even if it’s just for that sci-fi vibe. Creating a system that is both practical and useful is however far from easy, which is where Aedan Cullen‘s attempt at creating what he terms a ‘practical augmented reality device’.

In terms of requirements, this device would need to have a visual resolution comparable to that of a smartphone (50 pixels/degree) and with a comparable field of view (20 degrees diagonal). User input would need to be as versatile as a touchscreen, but ‘faster’, along with a battery life of at least 8 hours, and all of this in a package weighing less than 50 grams.

Glass Elephant

When looking at Aedan’s glasses-based prototypes with the single-eye AR experience, it’s hard not to flash back to the Google Glass hype of the past, beginning in 2013 with the release of the Explorer Version 1. These ran Android 4.4, and featured a 640 x 360 Himax HX7309 LCoS display, a 5 MP camera, 1 GB of RAM (2 GB for version 2). Interaction with the user took place both via the touchpad on the side of the device, along with a voice interface.

Despite Google’s push behind the device and accompanying media hype, Google Glass as a consumer device was a flop, not in the least due to the rather off-putting $1,500 price tag , leading to it being retargeted to the commercial market with the Enterprise Edition in 2017 and Enterprise Edition 2 in 2019. In 2023 Google discontinued the $999 Enterprise Edition 2, but a March 2022 teaser promised a follow-up product with a less ambitious scope. Meanwhile Microsoft’s HoloLens appears to be solid tech, but also priced outside the comfort zone of consumers.

Finally, Apple is rumored to be on the verge of releasing their own AR glasses, which would run RealityOS, offer about eight hours of battery life, among a whole host of yet-to-be-confirmed features. Against this backdrop it’s interesting to consider what a hobbyist can tinker together in their own workshop.

Off-The-Shelf AR

The AR prototypes so far.

To be  clear, what Aedan is showing off in the presentation is not a ready-to-use device that can compete with whatever Google Glass did and what Apple’s AR glasses may or may not do if they ever end up being released. What his efforts do show off very clearly is that even with modest means, a hobbyist can use off the shell components like the Sony ECX337A OLED microdisplay to get a very compact, power efficient and high-resolution (1280 x 960) image source for an AR device.

From ther,e much of the effort was with finding a way to optimize power usage, as the original STM32MP1 with Linux running on the Cortex-A7 side was a bit too power hungry even during idle at 1 W. This led Aedan to a power sipping, presumably more affordable SoC in the form of the Allwinner D1-based sun20iw1 SoC. This features a single XuanTie C906 RV64GCV core on a 22 nm process, and should use less power than the STM32MP1.

In the current iteration of Aedan’s AR prototype, the PCB has moved from the typical bulge on the side as seen with Google Glass to a ‘shelf’ on the front, which makes it protrude more from the forehead but definitely improves the profile when viewing the wearer from the side. It is assumed that as the PCB is further shrunk down, the size of this protrusion can be reduced.

The hardware and its Linux OS is capable of supporting a wide range of interfaces, making it an interesting platform to tinker with, especially with the target price of less than $500 for the whole package. Aedan mentions looking into eye tracking, for instance.

As Aedan himself puts it, even if no usable product emerges from these efforts, for him the biggest return have been the valuable lessons learned from miniaturizing and optimizing such a compact design.

10 thoughts on “Supercon 2022: Aedan Cullen Is Creating An AR System To Beat The Big Boys

  1. The advantage of the low price is that you have more money to spend on a barf bag and cleaning supplies. Another secret to a good design is that it can automatically call the paramedics when you get vertigo and fall and hit your head.

    This technology was invented by head trauma doctors who are looking for new revenue sources now that folks have figured out that football was designed to fill nursing homes with 40 year olds.

    1. I need some water…. That post was SALTY! If you can’t handle the motion.. don’t put it on. Probably an issue more with elderly people who won’t adopt the tech anyway.

  2. Interesting effort – but that’s not AR. The same as Google Glass was not an augmented reality device. It is a personal heads-up display.

    It maybe sounds as nitpicking but the difference is huge and very important. If you want augmented reality – i.e. to be able to overlay computer generated imagery over real world objects – and have it registered (= it “moves with the object it is attached to”), you need a very robust and accurate tracking system. He doesn’t have any at all, not even for tracking the position of the glasses in space.

    As a result you could display e.g. street navigation directions from your smartphone GPS but certainly not overlay e.g. maintenance information showing you which capacitor on a PCB you need to replace on a PCB in front of you.

    To be fair, even devices like Hololens are incapable of this – they are able to track their own position inside of a room using SLAM but can’t render information registered to an object without sticking e.g. fiducials to them. And then the user has to create their own tracking solution for this (which is complicated/nigh impossible to do because Microsoft’s SLAM drifts making rendering accurate overlays that “stay put” very difficult. Been there, done that …)

    AFAIK, Apple has been mum about this too – and given what they have in their ARkit for iPhone, it is very likely that their rumored headset will be able to track only itself but not the pose of objects it is looking at.

    (No, Microsoft’s spatial “anchors” and such won’t help you – they are only location “bookmarks” for a point inside of the SLAM space map but they won’t help you if you want to render an overlay on a box. The overlay would be referenced to the room and won’t stay with the box if someone moves it – the SLAM has no idea about the box at all).

  3. I’d almost wonder if using something like QNX as the operating system could reduce the requirements enough to be able to significanty improve over the (albiet potentially minimal) linux requirements

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.