"Loyalty, teamwork, cruel deception: welcome to robot evolution" via The Loom
I was trying to think of something snappier to write. But I couldn't. Damn these real scientists with real PhD's.
It's really cool to see a model of "random" change in process coding acted on by selection resulting in increased fitness. The various modelling of kinship realtionships on social behaviours in these robots was pretty eye-opening too.
Really it does bring up a whole heap of possibilities for evolutionary modelling.I think a very nice modification to the experiment would be to add either an environmental toxin "base" akin to the food one (except perhaps destabilising the robot for a few moments, which means less food points), or a predator robot population that counts the other robots as "food" (but perhaps restricted in mode of action - ooh "If you don't move they can't see you"). Bringing other robot populations (with different goals) I think would definitely be the most interesting step.