Evolution Sims

From WikiManual
Jump to: navigation, search

Our best to provide a comprehensive listing of simulations done in Darwinbots.

Before You Start

You must understand evolution before you can hope to judge the success (or lack thereof) of an Evolution Simulation (sometimes called an evosim). Below are some links to Wikipedia that provide useful knowledge that is generally unknown to the layman on evolution.

  • Punctuated Equilibrium - Theory that states that speciation and sudden evolutionary change can only occur in small gene pools. In real life, this tends to be at the edge of what the species finds survivable. These events are rare (hence "punctuated") and are separated by periods of little change.

Seems to be applicable in Darwinbots since the most striking mutations happen in regions where selection pressures are high.--Endychat 17:15, 24 Nov 2005 (MST)

  • Muller's Ratchet - Asexual creatures can accumulate "bad" mutations if mutation rates are higher than selective pressures. Eventually this leads to extinction.

Finding the Best

In Darwinbots and in real life there are those individuals who stand apart from the crowd. These are the founder mutants, the lucky few that have DNA that provides them a special advantage or interesting behavior.

Finding them though can be a challenge in and of itself. The easiest method is the Crown or King(add link) Button. A single click will highlight the "best" bot in a sim. In the beginning, many bots share virtually identical DNAs making this useless. As time progress however only those bots best adapted to the local environment will be selected, those that have passed the test of time and are more than fit for further evolution or in-depth DNA analysis.

Another method is to find good mutations by watching the success of their descendants. With a low enough mutation rate this is fairly easy, as the colors won't vary too rapidly for you to keep up with.

A final method is to simply look for those bots you find to be the most interesting. Perhaps they do an oddball dance, or use ties in an unusual way. Either way providing they can self reproduce and hunt well enough they can be scooped from their crowded native sim and dropped into a pristine simulation. These will then further develop their traits, with no boring brethren to worry about. Alternatively their DNA can be analyzed to find the source of their behavior, and the behavior incorporated into a more stable version.

Something else useful is the population graph provided in DB. Even while you're away with everything sped up, the graph will continue to faithfully display current population dynamics. Upon seeing a large gradual increase, you can take a look and find out how the population growth was made possible.

Once you find an interesting bot, you can then use the Save Robot DNA feature.

Your First Evolution Sim

Unfortunately, simply enabling mutations and waiting a few hours doesn't usually result in evolving complex behaviors. Using certain settings can greatly increase your chances of evolving an "intelligent" or interesting-behavior bot. Here are some general guidelines:

1: Don't be afraid to come by the forums and ask if you have a problem. Seriously, that's the reason why it's there.

2: Patience is key. The most successful evosims have usually run for over 1 million cycles and hundreds of hours.

3: Decrease the amount of veggies. If there are too many veggies, a bot will usually develop simple eating behaviors, such as just waiting in one spot for food to pass by. If there are just a few veggies, the bots will need to actively search for them. This can be achieved by going to general settings and decreasing maximum number of veggies or repopulation threshold to low numbers.

4: Increase sim size. The larger the sim, the more "intelligent" a bot will need to be to successfully find food.

5: Decrease Mutation rate. Although you may think the higher the mutation rate, the better, it's actually the opposite. If the mutation rate is low, then the good genes that the bot has already evolved have a higher chance of staying intact. You could be the best bot in the sim, but if your children are all mutated beyond any hope of survival, then it won't do you any good. A mutation rate of 1/32x is usually best.

6: Add shapes and edges. If a bot evolves in a complex environment, it will develop complex behaviors to deal with it.

Feel free to add to this list as you start making your own evosims! Good luck!

Commonly Seen Evolved Behaviors

Common Mutation Areas

These are all major locations that have been observed as likely to mutate, typically increasing or decreasing the amount a particular trait is expressed.

  • Continual Motion
  • fixed
  • Increased energy required for reproduction
  • inc/dec energy given to young at birth
  • continual shooting
  • easing of family avoidance
  • Deliberate family hunting
  • Use of tielen1 to affect birth tie
  • continual reproduction
  • Non-reproducing
  • Continual tieing
  • Increased body levels

Favorite Simulation Settings

Current evosims

See Moonfisher's evosims in the forum; includes a working Neural Network evosim, an advanced Zerobot sim, and a working co-evolution sim.

Also see the Co-Evolution topic, also in the forum, featuring ikke and Peksa's co-evolving evosims.