Hacker News new | ask | show | jobs
by KaiserPro 1692 days ago
> Augmented Reality at consumer level are already available, right?

yes, but not on battery power. As someone who is actively working in the field, there are about 5 unsolved problems before practical AR is a thing. (they are: location, display, object detection, audio understanding and power management.)

but this really isn't AR in the sense that a lay person would recognise.

1 comments

Yes, if with AR you mean "a system that recognizes the details of the environment and mixes relevant information in the visual field".

With «commercial products for Augmented Reality» I meant "systems that mix information to the environment in the visual field" - so, 'Augmented Reality Eyeset' (shortening as we were talking about optical displays).

--

Aside note: in actual AR, what is the issue you mentioned with «power management»?

> in actual AR, what is the issue you mentioned with «power management»?

if you take that headset linked in the level up, it has a power draw of 4.5 watts.

A practical AR headset, (ie one that you'd consider wearing in public, with normal people) has enough space for a battery with about 1.1 watt hours of power. In 5 years time that might be 2.4 watt hours(perhaps).

That headset will have about 30 minutes of light up time. Thats before we do graphics, sensors and all that jazz. So its basically not at all practical as anything other than a slight diversion.

So in order to get a practical headset style AR wearble system, you need a "puck" (I think thats magic leap's term for the CPU/GPU) to do all the heavy lifting, with a fair wedge of extra power. Now, that only solves one issue, CPU and or internet connectivity. You still need headset sensors.

This means multiple cameras for detecting position, and possibly another set for visual stuff (to allow projection mapping and object detection.) All of those cameras have either got to be low power enough and self contained (ie <=10mw and produce relative movement [or actual global pose] rather than picture data), or you've got to have a super low power way of connecting them to the "puck"

basically practical AR will require self contained cameras that can produce scene description(ie chair at position x,y,z) and cameras that produce absolute position (SLAM, but with a prebuilt map) all for a power budget of ~30mw or less.

I am very comfortable with the eyeset cable connected to the central unit in my pocket, battery equipped.

(And I instead feel very uncomfortable with the idea of a central unit near the skull... Have you considered that many could "feel" the same, and that it should be preferable to delegate much to a separate cabled unit also for market reasons?)