Code center > Darwinbots3
Realistic eyes?
Botsareus:
So I guess it is getting popular to post early stage DB3 ideas so I have one of my own:
We need more realistic eyes. I just don't see how seeing only ~7 robot lengths in front of you realistic even if we can adjust it.
Any ideas?
I have three. But will share only the most simple one and see where the thread goes:
How about we add a "second layer" set of eyes that see ~49 robot lengths ahead but only return basic logistical info?
spork22:
That might be cool.
Numsgil:
Here's what I'm currently thinking:
First, for some background, my current thinking is that each bot is composed of multiple "panels". This is how I'm going to do things around one bot engulfing another (ie: phagocytosis). Basically each bot is a simple polygon (possibly concave), and to control its shape it sets the internal angles of the polygon. The edges of this polygon are the panels.
Among other things, each panel can potentially sense light intensity. The light intensity is figured out using a simple raytracer (so it will have infinite range in theory, but since light intensity diminishes with distance actual range will be somewhat limited). So if you're deeper or in the shade, there's less light and you're less visible. Or if you bioluminesce, you're more visible. Also, things like shell could change your albedo, making you more or less reflective.
To detect motion, you'd implement something like The Reichardt Hassenstein Model in DNA code. Basically save out the light intensity values from the previous cycle and compare them to the current ones with some math. I might provide a mechanism to do this for you automatically, but I'm curious to see if it can be done in DNA, and if so, if it's interesting to do so. If you want to know distance, you'll need to do something with binocular vision. If you want a crisp image, you'd need a simple pinhole eye.
Obviously that only gets you very basic "I see something" or "I see something moving", but not really any details about what you're seeing. Which I partly sort of like: it makes being stationary a potentially advantageous defense mechanism. If you don't move, and you hide near an inert shape, and you have the same reflectivity as the shape, you're going to be very hard to detect. If I can give bots the important data like refshell, etc. through other senses, I might do it that way. Maybe through smell or touch. I'm still working through the details exactly, and I'm a little ways off from actually implementing any of it, so it may change as I think it through more. But that's what I'm thinking currently.
MysticalDumpling:
I like Num's idea.
spork22:
I do too. Actually, a lot of creatures do that in real life. They hide using larger creatures, which are more likely to be seen by a predator.
Navigation
[0] Message Index
[#] Next page
Go to full version