Adapting Morphology to Multiple Tasks in Evolved Virtual Creatures (2014)
Author: Dan Lessin, Don Fussell, Risto Miikkulainen


The ESP method for evolving virtual creatures [1] consisted of an encapsulation mechanism to preserve learned skills, a human-designed syllabus to build higher-level skills by combining lower-level skills systematically, and a pandemonium mechanism to resolve conflicts between encapsulated skills in a single creature's brain. Previous work with ESP showed that it is possible to evolve much more complex behavior than before, even when fundamental morphology (i.e., skeletal segments and joints) was evolved only for the first skill. This paper introduces a more general form of ESP in which full morphological development can continue beyond the first skill, allowing creatures to adapt their morphology to multiple tasks. This extension increases the variety and quality of evolved creature results significantly, while maintaining the original ESP system's ability to incrementally develop complex behaviors from a sequence of simpler learning tasks. In the future, this method should make it possible to build EVCs with complex and believable behavior.

[1] Dan Lessin, Don Fussell, and Risto Miikkulainen. Open-Ended Behavioral Complexity for Evolved Virtual Creatures. GECCO 2013.
Dan Lessin Ph.D. Alumni dlessin [at] cs utexas edu
Adapting Morphology to Multiple Tasks in Evolved Virtual Creatures Dan Lessin, Don Fussell, Risto Miikkulainen To Appear In Proceedings of The Fourteenth International Conference on the Synthesis and Simulati... 2014

Evolved Virtual Creatures as Content: Increasing Behavioral and Morphological Complexity Dan Lessin PhD Thesis, Computer Science Department, The University of Texas at Austin, Austin, Texas, December ... 2014

Neuroevolution Insights Into Biological Neural Computation Risto Miikkulainen To Appear In Science, 2024. 2024

Neuroevolution: Harnessing Creativity in AI Model Design Sebastian Risi, David Ha, Yujin Tang, Risto Miikkulainen To Appear In , Cambridge, MA, 2025. MIT Press. 2025