You are viewing bramcohen

Sun, Dec. 31st, 2006, 11:12 am
Sharks and Fish

The sharks and fish of wa-tor simulation is an interesting simulation of population dynamics, but has an extraordinarily unrealistic feature. It's way too 'geometric' - populations explode and die out within ridiculously small regions. This is mostly caused by the random walk strategy of movement of the creatures, which results in sharks eating all the fish in a very small region before starving to death.

This can be improved by making creatures pick one of the four compass directions and always head in that one until their death, although that can result in a shark dying because it's trailing behind another one. A way of improving on that, which just occured to me the other day, is for each creature to pick one of the four compass quadrants to head in (NE, NW, SE, SW) and then pick the ratio of the two directions it moves at random. On each turn, it decides which of the two based on whether it's currently above or below the goal ratio since it was born.

That should result in far less geometric and more realistic looking simulations. I haven't implemented this yet, if someone runs through the exercise please let me know.

Tue, Jan. 2nd, 2007 12:07 am (UTC)
akkartik: A slightly different idea

When a shark's been starving for a while (say it's down to a third of its food supply) it switches to 'migrate' mode. It keeps moving in a single direction until things improve (say it's at two-thirds full).

http://www.cs.utexas.edu/~akkartik/wator

Migration only works if sharks can live for a while without food, so I've scaled the default breed and starve rates by a factor of 10.

It takes a while to see differences with the original, but it seems like it makes the predators more efficient; over time the field supports more sharks and fish populations tend to stay more suppressed.

Tue, Jan. 2nd, 2007 02:51 am (UTC)
bramcohen: Re: A slightly different idea

The fish should migrate as well. It's their propensity to sit around like a bunch of oafs which causes them to die out so much.

Tue, Jan. 2nd, 2007 07:51 am (UTC)
akkartik: Re: A slightly different idea

Hmm. Fish don't starve so it's not clear what reason they would have to migrate. Do you mean fish should migrate away from predators? I don't think that is realistic.

It's not clear to me that having fish 'die out so much' is undesirable. Perhaps I'm not as good at reading the evolving patterns or judging their realism. Do you think your original idea is better than migration in this respect?

Sun, Jan. 7th, 2007 06:34 am (UTC)
bramcohen: Re: A slightly different idea

The reason for fish to migrate is to migrate away from each other, to avoid forming a big juicy area for sharks to feed on a whole bunch of them at once. Also to give them space to breed into - they frequently don't breed because literally the space isn't available.

I think I see what you're trying to accomplish by having sharks move randomly when they're well fed, but it would be better to make them simply make 90 degree turns whenever they feed. Making fish migrate would make having shark stay in place at all be completely pointless.

Thu, Jan. 4th, 2007 06:54 pm (UTC)
mongoltrophies: totally offtopic; apologies.

I don't know if you've heard of this software called Pando, but it seems very similar to bittorrent, only they employ their own server to help seed, which they pay for by presenting ads in the client software. In their own words: "Pando's patent-pending P2P-Web hybrid platform combines the efficiency of BitTorrent with the reliability of a robust data center." (from here)
Is it legitimate for them to try to file a patent for this, which appears to be just an implementation of bittorrent (and yousendit)?