Bots and Simulations > DNA - General

Feeding and body

<< < (2/3) > >>

EricL:
Today, an nrg shot goes 80% to nrg, 19% to body, 1% to waste.

Reproduction in evosims depends upon both body and nrg.  Bots must have at least 2 body to reproduce.  

We should be aware that completely separating the aquisition of nrg and body would require bots to evolve both an nrg aquisition strategy to fund their activities and a body management/aquistion strategy to replenish enough body to be able to reproduce.  That is, in my expereince, it is extremely common for simplisitc evobots to sit at body <= 2, attempting reproduction every cycle and for the (indirect) aquisition of body through nrg feeding to be the main determinant of reproductive sucess.

Evolution needs a gradual slope leading from one adapation to another, ideally a single mutation be in a single bp or larger sequence.   It strikes me as a quite large chasam for evobots to cross to require that they must evolve body management/aquisition in addition to nrg aquisition in order to sucessfully reproduce.  For this reason, while I would not be opposed to changing the above percentages, I think I am against a complete separation.

Numsgil:
It's really a tough issue.  As it is right now, body management is more icing on the cake.  The program does enough for the bots to survive without being too sophisticated.  If you totally remove it, it really hampers evo bots as Eric pointed out.  Simpler bots, like animal minimalis, would also stop working.

Maybe we can remove nrg shots giving body, but change it so when nrg "overflow" past 32000, it's automatically converted to body (this might be true for some parts of the program, like veggy feeding, but we could try standardizing it so that this is always true).  That removes the automatic body growth for bots that aren't really trying while still giving a way for bots to acquire body that don't specifically code for it.

However, the reverse shouldn't be true.  When a bot reaches 0 nrg, it should die, regardless of wether it has body left or not.

Testlund:
If you would be interested to give this a try, I would be happy to set up a new evosim on my spare computer to run it around the clock. Maybe a combination of both, like when reaching 32000 energy it will start to get converted to body or if a bot evolve the sysvar to convert energy to body it can use that sooner, getting an advantage over the other bots. I'll jump at everything that makes DB closer to realistic metabolism.  

Peksa:
I don't really like program set limits like nrg shots returning body after 32000 energy. I prefer Eric's more gradual way to limit body gain, ie. setting the body gain percentage much lower. 1-5% should be low enough. I personally feel that lower the better.

By the way, what happens now to nrg shots bot's hit with if it has 32000 energy?

If it is lost (or even more so if it's turned to waste) it could encourage, even if very slightly, evolving some kind of body management. Though I wouldn't think that effects of having 32000 are high concern for evolving bots

Numsgil:

--- Quote from: Peksa ---I don't really like program set limits like nrg shots returning body after 32000 energy.  I prefer Eric's more gradual way to limit body gain, ie. setting the body gain percentage much lower. 1-5% should be low enough. I personally feel that lower the better.
--- End quote ---

That's the other option.  I still think it's better to have body gain be 0 from nrg shots, and then give bots body when they pass 32000 nrg as part of the nrg overflow, instead of just tossing the nrg away.  The conversion could even happen at a substancial loss, since it's an overflow method.


--- Quote ---By the way, what happens now to nrg shots bot's hit with if it has 32000 energy?

If it is lost (or even more so if it's turned to waste) it could encourage, even if very slightly, evolving some kind of body management. Though I wouldn't think that effects of having 32000 are high concern for evolving bots
--- End quote ---

 

It's just discarded at the moment.

Navigation

[0] Message Index

[#] Next page

[*] Previous page