Evolving virtual creatures in Lenia

Lenia is a cellular automata capable of producing lifelike and interesting virtual creatures in 2D. The patterns produced by Lenia are self-organizing, and are created by relatively simple rules. Though the creatures that can be produced in Lenia are diverse, looking through the space of creatures to find the interesting ones is time consuming, as many of the creatures are similar to each other. We would like to automate the search for novel and interesting creatures.

Bildet kan inneholde: blå, azure, gjøre, elektrisk blå, sirkel.Bildet kan inneholde: blå, azure, gjøre, elektrisk blå, sirkel.

Images created with Lenia[1]

In evolutionary algorithms it is common to optimise for an objective defined by the algorithm’s fitness function. When looking for creatures in Lenia we are not interested in creatures that fulfill a predefined objective, rather, we would like to discover a wide array of different behaviours. To achieve this goal we would like to turn to quality-diversity or open-ended algorithms, in which novelty and interestingness play a large role.

Here are some examples of directions you could take within this project:

  • Designing your own open-ended algorithm that produces a repertoire of interesting creatures in Lenia.

  • Evolve creatures in Lenia with several existing open-ended algorithms. Compare the algorithms by analysing the creatures that were found.

[1] Chan, Bert. (2019). Lenia: Biology of Artificial Life. Complex Systems. 28. 251-286. 10.25088/ComplexSystems.28.3.251. 

[2] Bert Wang-Chak Chan; July 13–18, 2020. "Lenia and Expanded Universe." Proceedings of the ALIFE 2020: The 2020 Conference on Artificial Life. ALIFE 2020: The 2020 Conference on Artificial Life. Online. (pp. pp. 221-229). ASME. https://doi.org/10.1162/isal_a_00297

[3] Kenneth O. Stanley; Why Open-Endedness Matters. Artif Life 2019; 25 (3): 232–235. doi: https://doi-org.ezproxy.uio.no/10.1162/artl_a_00294

 

Publisert 18. aug. 2021 17:13 - Sist endret 19. aug. 2021 12:57