GECCO 2017

IDG members Daniele Gravina and Antonios Liapis attended the Genetic and Evolutionary Computation Conference in Berlin, presenting original work from the Institute of Digital Games on evolutionary divergent search and procedural content generation. Daniele Gravina presented work co-authored with Antonios Liapis and Georgios N. Yannakakis titled Coupling Novelty and Surprise for Evolutionary Divergence which combines two divergent search algorithms (one trying to find novel solutions and one trying to find surprising solutions) and tests their performance in a maze navigation task. Daniele also presented a poster on evolving surprising soft robot morphologies, summarized in Exploring Divergence in Soft Robot Evolution. Find more about surprise search here!

On the other hand, Antonios Liapis presented cutting-edge work on dungeon generation where evolution can be applied either to generate the rough outline of the dungeon and its connectivity, or to generate each segment within the dungeon with all the game details such as monster and treasure placement, or both! The research is summarized in the paper Multi-segment Evolution of Dungeon Game Levels and allows for a fast and controllable way to generate dungeons of any scale based on techniques popular in games such as Diablo and Spelunky: compartmentalizing the generative task into segments connected together as jigsaw pieces.