Soft Robot Evolution with Surprise Search

Divergent search is a recent trend in evolutionary computation that does not reward proximity to the objective of the problem it tries to solve. This can be particularly useful in domains such as games or art where the space of possible solutions may be deceptive or where the best solution may be subjective. A particularly deceptive domain is the evolution of soft robot morphologies, i.e. robotic structures with distributed deformations which can reach any point in the space with an infinite number of configurations (in theory). This project tests the robustness and efficiency of divergent search algorithm such as MAP-Elites, novelty search, surprise search, and their combination in this challenging testbed.

Relevant Publications

  • Daniele Gravina, Antonios Liapis and Georgios N. Yannakakis: "Fusing Novelty and Surprise for Evolving Robot Morphologies," in Proceedings of the Genetic and Evolutionary Computation Conference, 2018. PDF BibTex

  • Daniele Gravina, Antonios Liapis and Georgios N. Yannakakis: "Exploring Divergence in Soft Robot Evolution," In Proceedings of the Genetic and Evolutionary Computation Conference, 2017. PDF BibTex

  • Daniele Gravina, Antonios Liapis, Georgios N. Yannakakis: "Blending Notions of Diversity for MAP-Elites," in Proceedings of the Genetic and Evolutionary Computation Conference, 2019. PDF BibTex