Robot Choreographies: Artificial Evolution between Novelty and Similarity

In this paper we introduce a novel fitness function for evolutionary art, which generates sequences of movements--i.e. robot choreographies--based on similarity to an inspiring repertoire. Similarityis counterbalanced by a novelty mechanisms, which makes it possible to sample unexplored areas ofthe choreography space. The approach is discussed together with preliminary results achieved in thecontext of Nō theatre. This work is a first step towards the development of a computational creativitysystem that can incorporate diverse generative mechanisms and can exploit information theoretic andcomplexity measures both for the generation and assessment of the choreographies produced