Learning behaviors from a virtual agent

Authoring complex behaviors for robotic agents has always been prohibitively expensive. Inspired by the Restaurant Game, this project seeks to explore the idea of data-driven behaviors, sourced from a crowd of people that act as if they are taking on the social role of a robot in a mock up cooperative task. After collecting these new behaviors from the internet, our robot, Nexi, utilized these situated actions in a real world task, interacting with everyday users at the Boston Museum of Science.