I recently rediscovered DB, and decided to try find a better way of getting an evosim running. After a few attempts, I got a pretty good system running, and got results much faster than any other way. Basically, the environment is made up of a chain of simulations, each connected to the next and set up to evolve a certain kind of behaviour. For example, the initial simulation would have veggies that both feed and force reproduction in the target cells. The cells are all clustered in one spot, with an out-teleporter leading to the next simulation so that only mobile bots can proceed. In this next simulation, the veggies still feed the bots, but do not get them to reproduce. Usually the mobile bots in the previous simulation reproduce more frequently (more area covered), so a steady supply of bots are available to evolve. This would then have a third out-sim to a simulation where the veggies do not feed at all or very unreliably, so to encourage feeding behaviour. This method allows some control over the evolution path of the bots, and I have gotten one or two simulations after a few hours that display some form of stable reproduction.