Google improves AR by 89%
Back in my day (circa 2015), we used to have to send Google ADFs (Area Description File) through email and backload them into our apps, crossing our fingers that those fear-inducing 10 seconds of app-freezing load time wouldn’t crash Tango Core, all to sync up our experiences in real world space.
Then over the years came AR Cloud companies like 6d.ai, YouAR, and Fantasmo, which gave promise to the notion that ubiquitous visual positioning and AR multiplayer were only a horizon’s step away…
… And they were right, because yesterday Google came roaring back with Cloud Anchors, a smoothed & polished reintroduction of ADF and possibly the first true public foray into the AR Cloud eventuality.
To highlight the immense progress Google’s made in 3 short years, here’s a comparison between ADF-based syncing and Cloud Anchor syncing:
ADF
- 3 minutes to scan, export, and manually email the ADF file
- 2 minutes to open email and save ADF file
- 1 minute to open ADF file and scan space to sync
= 6 minutes to sync with ADFs
Cloud Anchors
- 15 seconds to find plane and place anchor
- 15 seconds of manual typing to connect to host
- 10 seconds to scan and sync
= 40 seconds to sync with Cloud Anchors, an 89% improvement
Final Thoughts
Beyond UX and setup time improvements, what makes Cloud Anchors truly special is for the first time developers can build collaborative, cross-platform augmented reality experiences right out of the box, though there are shortcomings that must be mentioned:
Syncing failure: We tried out various tabletops, wood floors, rugs, carpets, and tiles but most surfaces we tried failed to sync, seemingly at random. We instead retreated to the high contrast ideal of red carpet on wood floor, which worked perfectly even from extreme angles and distances.
Going forward, ARCore will need to include a validation step for Cloud Anchor data or users will become frustrated and App Store ratings will suffer.
Syncing accuracy: When you’re lucky enough to actually sync, the accuracy of the anchor placement can be off by as much as 5cm. This may seem negligible, but many AR experiences (like ours) need sub centimeter accuracy.
Troubleshooting Tips
If you’re reading this, you’re probably a developer or Xcode-capable enthusiast, so here’s some helpful tips from us:
- If scanning doesn’t seem to work, turn your phone’s wifi off and on. We ran into critical syncing/connection problems twice and this fixed it.
- Ensure the Cloud Anchor was placed, created, and saved to the cloud before any clients attempt to connect to the room.
- Place your anchor in a location that has unique edges and points with varied colors, but not overly busy. Stay away from repeating patterns.
- The room # and IP Address have to be typed in manually by the user every time, so we recommend using Firebase or Photon Bolt to automate this.
(We will continue to update this list with new Cloud Anchor-related tips so if you have any yourself, be sure to let us know!)
Hi, I’m Aidan Wolf, a pursuant AR dev and storyteller trying to figure out the better ways to tell immersive stories in the real world.
Follow me on Twitter at @Aidan_Wolf to keep up with our projects!
Bye for now! 👋
