Darwinbots Forum
Code center => Bugs and fixes => Topic started by: Testlund on December 12, 2006, 04:43:30 AM
-
I find it strange that my bots that have evolved for weeks just got bad suddenly and keeps dying nomatter how many times I keep restarting the sim. Started a save where the bots had become 1000 but they all die. I was wondering if Eric had done something about how costs are being applied in one of the later buddy drops, maybe 2.42.9j. Is it really possible that EVERY bot just happend to get bad mutations? The highest mutation rate I have is in point mutation and that's set to 2500 and the bots multiply quite often, which means there should allways be some good ones to keep up the population after all this time. It's like evolution just went backwards.
I have a sim here with 1000 bots you can check out. Note that morphological cost for rotation is set to 0.115, but I don't think that new cost is what is causing the bots to die. I have tried with a value set to 0.01 too and most bots doesn't rotate that much. When you start this sim the bots should all be dead within 20000 cycles.
-
Ooh a veg sim, at not even 1 cycle/sec I imagine it will take a while to reach the 20000 cycle limit you suggested but I'll leave it a while and see if anything happens.
-
Thanks, Jez. You could also let it run overnight.
-
I would bet a good beer that its #59) that did it. Negative indirect memory references used to get dropped. Now they are abs(n) mod 1000.
Your evolved DNA now does something differnt than it used to. Sorry.
-
May I pour that beer over your head, Eric?
No, seriously... A bug or shortcoming should allways be fixed nomatter the result I think. Don't want bots to take advantage of bugs. Fortunately I had a bot which I saved way back when the sim started to get interesting so I started a new sim with that. It seems to fare better. It's been running for allmost 15000 cycles now (5000 cycles with night) and there are still 800 bots left.
-
You could probably salvage even the most recent bots if you deleted all base pairs of the form *-n or -n *from their DNA...
-
Looks like I need to do just that cause the bots I'm running now are close to be whiped out too, though it looked promising first. I'll see if I can find that stuff in the DNA.