Towards a Human-in-the-Loop Framework for Swarm Intelligence-Driven Dance Generation

This paper introduces a preliminary implementation of the Feature Mapping module, a core component of the novel approach that exploits swarm intelligence algorithms to generate dance choreography by virtual agents. We delegate the mapping process from music to simulation parameters to a neural network, casting its development as an optimization problem. Additionally, we propose an initial qualitative evaluation apparatus in the form of a questionnaire to assess the performances generated. We leverage preliminary human evaluation results to assess the automatic evaluation system employed, specifically the fitness functions and hyperparameters used. This work represents a first step towards the realization of a human-in-the-loop framework for the generation of artistic performances. Moreover, this leads to the realization of eXtended Reality application scenarios.