Hey Numsgil, think about this picture. This was stuck in side my head for a while. I think it may be possible.
So I will implement my 'workaround' and make sure it works the same as it did in my 2.44 experiments.
I wait until non mutating species die as I do now.
Then, instead of running 'find best' as I do now, I will run the following algorithm:
Repeat 15 times:
Find best * genetic distance -> save this robot, remove all robots from the search query within 5% of genetic distance.
Now I got 15 reasonably unique robots (although can be less if I run out of query)
Instead of restarting with one specie like I do now, I restart with all 15 (they will be all named the same but within different folders so the program internally handles them as one specie) so there will 30 total robots (hopefully producing there own chloroplasts at some point, therefor 2 species)
Cool, there is your basic model for eco-system evolution.
Then we can add Internet mode where all users can run this simulation with different settings (global controllers (my workaround) will locally remain the same by checking if the global settings mach in the .dbo file being transferred)
And wala, you got eco-system evolution.
I am also looking into (when I get to it) If I can start 'eco-system' evolution from a zero bot.