Smart adventurers wear their eyes on the outside . . .

[Anthony]

Originally Posted by jeff_wilson

Why not? Each puff of surveillance dust can see as well as a human eye, apparently.

We'll call this 'magic'. Minimum size for human-equivalent vision is about a 2mm lens.

[dcarson]

Can you use multiple inputs to make a virtual lens big enough? That's how the VLA radio telescope works. The amount of detection strength is based on the area of the elements but the resolution is based on the size of the distance between all the elements.

[jeff_wilson]

Originally Posted by dcarson

Can you use multiple inputs to make a virtual lens big enough? That's how the VLA radio telescope works. The amount of detection strength is based on the area of the elements but the resolution is based on the size of the distance between all the elements.

That comes under " phased arrays or interferometers", but I'm not sure why Anthony considers them unlikely.

[Anthony]

Originally Posted by jeff_wilson

That comes under " phased arrays or interferometers", but I'm not sure why Anthony considers them unlikely.

Because you need extremely accurate and consistent spacing between your optical elements to do that, and air motion is not your friend. A group of crawlers that could park, link up, and form an optical interferometer is more possible.

[jeff_wilson]

Originally Posted by Anthony

Because you need extremely accurate and consistent spacing between your optical elements to do that, and air motion is not your friend. A group of crawlers that could park, link up, and form an optical interferometer is more possible.

IIRC, the positioning for interferometry needs about 1 part in E+8 or E+9 accuracy; the birthday paradox allows this to be achieved a majority of the time with a cloud of <100k motes.

[Anthony]

Originally Posted by jeff_wilson

IIRC, the positioning for interferometry needs about 1 part in E+8 or E+9 accuracy; the birthday paradox allows this to be achieved a majority of the time with a cloud of <100k motes.

No it doesn't -- because you need to know which mote is correct.

[sir_pudding]

Originally Posted by Anthony

No it doesn't -- because you need to know which mote is correct.

Can't you do that with processing?

[jeff_wilson]

Originally Posted by Anthony

No it doesn't -- because you need to know which mote is correct.

What have you done with the real Anthony? Obviously each of n motes checks with its n-1 siblings for outputs that can produce useful images. Even if you need two siblings with the same distance from a third so that you get 2D images, that's only ~=n*n combinations. You almost certainly get one useful triplet and usually more.

[Anthony]

Originally Posted by jeff_wilson

What have you done with the real Anthony? Obviously each of n motes checks with its n-1 siblings for outputs that can produce useful images.

The problem is, there's no statistical way of doing that. The requirement is that you know the distance to the other mote to within a fraction of the wavelength; as long as you know that, you can produce useful images with any mote, and if you don't know that, you can't produce useful images ever.

[ldj00]

Hello,

Without accurate position and direction facing sensors and computation, it would never be useful, so it has to have both.

If they are communicating wirelessly to each other, then I am assuming some mesh network, they could time the signal from their neighbors and compare that timing. That should allow each small group to accurately position those neighbors...

Seems feasible with current existing products, but just not at that size (Raspberry Pis). Assuming dedicated circuitry, I do not see why it would not be feasible at dust-mote size...