Code center > Suggestions

Suggestion about fixed bots

(1/2) > >>

Testlund:
I think bots shouldn't be able to appear on top of each other which makes several hundred bots look like one bot. Even if they are fixed they should still bump away from each other or at least only stay very close and form colonies. Anyone agrees?

EricL:
The whole concept of a fixed bot is kind of weird IMHO and is inelegant and unrealistic from a physcis perspective (so are moving shapes BTW).  As an example, fixed bots are treated as if they mass 32000 for the purposes of collision elasticity.

Fixed veggies used to be important for topological reasons, creating poor-man's walled nursuries for zero bots for example, but shapes can fill that role now.

Having fixed veggies is interesting as a means of creating topological separtion and thus species isolation in large evo sims, but since fixed bots can be unfixed, this is only so effective for so long.  Fixed veggies get unfixed after a few million cycles for example in my southpacific sim as soon as some bot mutates to shoot the right memory shot.  Were I to do such a sim again, I would use shapes to acheive topological isolatiuon.

About the only thing I like about the whole concept of being fixed it is that it enables a class of bots which use it as a proxy for gripping and ungripping the surface as a means of locomotion such as Inchworm.

Perhaps what we need is to do away with the whole concept of fixed all togther and replace it with a mechanism that allows bots to change their coeffecient of static friction as a means of rooting themselves.  This would allow for rooted plants and inchworm style bot motivation, but such bots could still be displaced soem ways by a large enough impact.  Note that my vision for shapes is that they too will have mass proportional to their size and that it will be possible for the shape itself to be the one deflected...

Anyway, in the near term, I could probably make it so that fixed veggies separate until they don't overlap when a shape sweeps them or when they materialize or teleport on top of one another.  Moving shapes would still sweep them into clumps, but the clumps would not overlap, at least not for long.

Numsgil:
Oops, I wrote up a post about this just now in another thread   linky

It solves the problem of overlapping, and I think that's how the physics in the C++ fork works.

As to wether the whole concept of ties makes any sense at all, I agree some sort of "griping" function is a good idea and can largely replace being fixed as it is now.  Some caveats though: if we go into 3D, bots would need to be against a wall or shape for this to really make any sense.  In 2D it can be everywhere if we imagine the bots as marbles on a pane of glass, but doesn't make much sense if bots are in a fishbowl, and the world is a side on view.  In those cases, fluid dynamics would be the thing to exploit.  Gigantic light weight bots are going to have a great deal of added mass, which should make them increasingly "fixed".

Make of that what you will

Sprotiel:

--- Quote from: Numsgil ---Gigantic light weight bots are going to have a great deal of added mass, which should make them increasingly "fixed".
--- End quote ---
Did I ever say that "added mass" was a physically unrealistic and silly concept? If not, that's an oversight corrected. However, changing your sentence to "Gigantic light weight bots are going to experience a great deal of drag, which should make them increasingly "fixed"." saves your argument, so I'm just nitpicking.

Anyway, EricL's plan sounds fine, though I'd rather consider shapes as walls and don't care much about they should move.

Numsgil:
Added mass isn't an unrealistic concept at all.  Just because you've never heard of it before...  I wrote the wiki article on it, feel free to go check it out here.  The talk page has a nice dialogue between me and another naysayer that should convince you that I'm not just making things up

Navigation

[0] Message Index

[#] Next page