Author Topic: Reccomendations  (Read 4040 times)

Offline falco1029

  • Bot Neophyte
  • *
  • Posts: 4
    • View Profile
Reccomendations
« on: February 08, 2008, 08:28:14 AM »
The main reason I got this was to see if I could have it simulate some actual evolution in logic among the bots, but whenever I do a simulation, the mutations just cripple the bots, never really giving any advantageous mutations. I generally use Animal Minimalis and then Alga_Minimalis for food. Any suggestions on some good settings for maximizing evolution?

Offline Endy

  • Bot Overlord
  • ****
  • Posts: 852
    • View Profile
Reccomendations
« Reply #1 on: February 08, 2008, 09:02:33 AM »
The easiest is to just alter the physics in the enviroment. Turning gravity to at least Earth level creates an interesting waterfall effect, you can also turn off torroidal mode to create a pit. Shapes are also an option, giving the bots a barrier they have to figure out or a platform to wait on.

Swapping Alga_Minimalis could help too, without a decent prey the bots don't have much of a challenge. Alga_Min also tends to repop more often than not, keeping it from evolving.

The Alga below is good for seeing some extremly rapid mutation. They're near impossible to wipe out so they don't repop so often. There are other veggies out there designed to act like prey, adding one of those can also help to spur the bots on.

' AlgaRepro
cond
start
99 .mrepro store
314 rnd .aimdx store
'duplicate code below as many times as desired
32000 rnd 999 rnd store
stop

Adding an old age cost can also help with population turnover. Getting rid of the older bots can help to clear the way for younger bots to have a better shot.

Offline falco1029

  • Bot Neophyte
  • *
  • Posts: 4
    • View Profile
Reccomendations
« Reply #2 on: February 08, 2008, 09:06:55 AM »
Alright, I added that algae instead of the other one and also added this p,ant that runs away from predators that I found in the beastiary. I guess I'll add some barriers and maybe some gravity as well. Thanks.

Offline shvarz

  • Bot God
  • *****
  • Posts: 1341
    • View Profile
Reccomendations
« Reply #3 on: February 08, 2008, 10:58:17 AM »
One note of caution: Evolution is pretty difficult to spot even when it happens because bots figure out ways to be better, which are not obvious to us.  

A common way to test for it is to run a competition of the evolved bot and the one you started with.
"Never underestimate the power of stupid things in big numbers" - Serious Sam

Offline Testlund

  • Bot God
  • *****
  • Posts: 1574
    • View Profile
Reccomendations
« Reply #4 on: February 08, 2008, 12:55:15 PM »
Has anybody managed to evolve a better bot from a predesigned bot? It has been mensioned this doesn't work well because any mutation will just break such bot giving it a disadvantage to the other ones. The best way is to start with a zerobot or randombot. Then a bot that doesn't do anything can only get better. Start a mutation with only morphological costs and perhaps a small body upkeep cost, then add an ageing cost as soon as you have a replicator.
« Last Edit: February 08, 2008, 12:57:34 PM by Testlund »
The internet is corrupt and controlled by criminally minded people.

Offline EricL

  • Administrator
  • Bot God
  • *****
  • Posts: 2266
    • View Profile
Reccomendations
« Reply #5 on: February 08, 2008, 01:36:34 PM »
Mutating bots very often do much better than their equivalent non-mutating brethern in internet mode.  This happens often enough and consistantly enough that I have been forced to modify my position that hand authorred bots represent a steep, otherwise unreachable pinicale in the evolutionary landscape and that they must devolve before they can evolve (this is sloppy terminology, but you get my drift).

The thing I've come to realize is that the morphospace is not only vast but vastly multidimensional.  Thinking about it as a 3D space with peaks and valleys is misleading.  Hand authorred bots may represent a steep pinicale in many evolutionary dimensions, meaning that a single mutation in that direction breaks them and lowers their fitness to such an extent that they cannot compete, but there are so many dimensions, so many different genotypes adjacent to a given genotype in the space of all possible genotypes, that even the most fragile hand authorred bot can still increase it's fitness via a mutation in at least some directions.



Many beers....

Offline falco1029

  • Bot Neophyte
  • *
  • Posts: 4
    • View Profile
Reccomendations
« Reply #6 on: February 08, 2008, 03:04:13 PM »
Alright, I'll see about all of that. I just like seeing something like evolution occuring in digital organisms on a personal computer. It's quite an interesting phenomena and I love this program because of it.

One quick question though. It seems if I restart a simulation it saves some of the mutations to the original bot. Am I imagining things or is there an option I need to turn off?

Offline EricL

  • Administrator
  • Bot God
  • *****
  • Posts: 2266
    • View Profile
Reccomendations
« Reply #7 on: February 08, 2008, 03:16:29 PM »
Quote from: falco1029
One quick question though. It seems if I restart a simulation it saves some of the mutations to the original bot. Am I imagining things or is there an option I need to turn off?
If you reload a sim file (I.e. do a File->Load Simulation) it will restart that sim file from the point where it was last saved to that file.  Any mutations, deathes, births, etc. that may have occurred in the runnign sim that are not saved will be lost at that point.  If you simply quit and restart the program, it will save the currently runnign sim at exit time and resume that last sim you were running from the point at which you quit when you restart.  That is, quitting DB saves the state of the currently running sim to lastexit.sim.  Starting DB will automatically load and start (or as 2.43.1, offer to start) the lastexit.sim automatically.
Many beers....

Offline falco1029

  • Bot Neophyte
  • *
  • Posts: 4
    • View Profile
Reccomendations
« Reply #8 on: February 08, 2008, 03:28:02 PM »
Ah, alright, thanks.