[piva00]

Isn't construction/architecture nowadays moving more and more to BMI software?

I'd assume that with a BMI model and AR capabilities (even from a phone) where you set a square (or some other shape) with known dimensions on the ground to calibrate and use GPS (or other positioning systems) for determining position/heading, doesn't seem like software for visualisation of models in-site should be that hard. Even on a current phone it should be doable.

What am I missing? Because it feels like there is some complexity I'm not considering.

[AndrewKemendo]

It would be too much to really go into, but what I will say is that almost all of your assumptions are misguided.

I don't mean that in a mean way, but for example using GPS for AR anchoring is not viable and will never be. So we have to build computer vision systems for localizing content, and those are complex and expensive and have hardware, bandwidth and other constraints.

BIM is ubiquitous, but not easily integrated with other hardware, and the hardware best suited for it (HMD) are expensive and not yet everywhere - or even widely adopted.

[fsloth]

"GPS for AR anchoring is not viable and will never be"

You can get centimeter level accuracy from GNSS with off the shelf custom antennae and software, see e.g. Trimble Catalyst https://catalyst.trimble.com

You can combine that with AR devices to get good enough positioning onsite for review purposes.

It's not precise enough for attaching nails and such, for which there are other location based solutions.

We actually have most of the pieces for a 'smart construction site' in existence as off the shelf components. But the construction industry is such that startups are not very well positioned to lead the change - hence I understand you disenchanted tone.

[AndrewKemendo]

Well GNSS =/= GPS, so my point stands, but...

No, GNSS is not ready to go. Look at the submeter coverage area for GNSS, it's TINY. Go look at work that Todd Humphreys has been doing at UAustin for over a decade trying to do cm specific GNSS localiztion with wider area deployments. It's just not tractable, scalable or anything near ready for a mass market. Like there is no reason for GNSS to grow at all when we're now building cloud vision based localization systems.

You keep referencing Trimble, I get it, they are doing good work in a handful of these areas. But the reality is that the uptake of hardware by firms to support AR at a scale that would support a company, is nowhere close to ready.

I'm telling you, I've skinned this cat from every conceivable angle.

[fsloth]

"But the reality is that the uptake of hardware by firms to support AR at a scale that would support a company, is nowhere close to ready.

I'm telling you, I've skinned this cat from every conceivable angle."

Ok, that angle I can agree on. I don't think there is any way a small company can provide an AR component and hit it big.

However, if an AR solution is bundled next to other computing infrastucture, with the import of files etc. handled, I've not seen anyone dismissing this concept.

The problem is that you need the entire infrastucture for project management and engineering to be digitized, and then you can add up the AR component to that.

Then AR is just one more "check box" in the feature lists of application suites offered by the incumbents. I think it will become an important checkbox, but like you said, there won't be an "Uber of AR" like there is no "Uber of DWG import".