Series of papers on evolving morphology and control systems for virtual creatures.
- Goal is focused on controllability and total complexity of behavior
- Approach is to break creation into multiple distinct stages
- creation of morphology & initial control
- train multiple new controllers for multiple small, distinct tasks
- train a combo controller to select and switch between individual controllers (pandemonium)
Dissertation pulls all the papers together
[Open-Ended Behavioral Complexity for Evolved Virtual Creatures – GECCO 2013](http://itu.dk/people/dles/gecco_2013.pdf)
[Adapting Morphology to Multiple Tasks in Evolved Virtual Creatures – ALIFE 2014](http://itu.dk/people/dles/alife_2014.pdf)
[Trading Control Intelligence for Physical Intelligence: Muscle Drives in Evolved Virtual Creatures – GECCO 2014](http://itu.dk/people/dles/gecco_2014.pdf)
[Soft-Body Muscles for Evolved Virtual Creatures: The Next Step on a Bio-Mimetic Path to Meaningful Morphological Complexity – ECAL 2015](http://itu.dk/people/dles/ecal_2015.pdf)
Something is missing in evolved virtual creatures work by Lessin -- and I think that's an ecology. Just evolving a creature on a plain plane isn't very compelling, as it's just locomotion, not life.
Ie, controllability isn't the interesting part of complex behavior.