Creative Robotics

The Performative Body-Mapping (PBM) method develops a novel approach to designing social, non-humanlike robots, which tackles the ‘correspondence problem’ in demonstration learning by harnessing dancers’ movement expertise and taking advantage of their ability to embody unfamiliar forms. Motion capture of performers inhabiting costumes that resemble robots provides the necessary data to bootstrap a machine learning process, allowing non-humanlike robots to learn how to move in ways that are readable to humans. Current research is exploring the combination of movement models learned using PBM with intrinsically motivated machine learning to support the improvisation of human readable movements, expanding the expressive range of non- anthropomorphic social robots.