DeepCraft: Co-Intelligent Architecture and Human and AI-Driven Craftsmanship in Design-to-Production Pipelines

Abstract

The working paper investigates the potential of artificial intelligence technologies (AI), namely the Generative Adversarial Imitation Learning (GAIL) implemented in a process of digital robotic fabrication prospectively to be used in craftsmanship. The method introduced is based on a preliminary demonstration provided digitally in an abstract toolpath generated by a human-driven movement in a hand gesture translated into a digital space in a real-time process. The investigation presented in this paper focuses on a preliminary computational digital framework which may serve as a base for further investigation. At this stage of the report, the framework encompasses human hand recognition creating a toolpath for a robot, which learns its principles and tries to interpret the process in a digital space. This learned toolpath resulted in a digital brain being applied again in a different shape of the human-created toolpath or gesture movement. The paper also presents the computational system of the real-time navigation of the robot based on a human gesture in a virtual space. The learned knowledge by a robot is observed in a digital environment before any physical applications.