Reinterpretation of Gabo’s Constructivism through AI Deep Learning & Robotic Fabrication

Starting from Naum Gabo’s Linear Construction sculpture series, creating the ruled surface between curved rail defines the geometry that we can analyze and utilize. Digitally reproduced sculpture extracts certain fundamental logics and signatures of the geometry into an element. After we analyze this distinct recognizable and reproducible element from the sculpture, we began to feed images of the element as latent space into convolutional neural networks (CNNs), GANs (Generative Adversarial Networks). Building digital models with identifiable visual characteristics, we can feed images into a neural network to generate blueprints from noble spatial relationships. Specifically, we structure explicit visual signatures as embed information. With a ruleset, taken from Naum Gabo’s sculpture's analysis, we can extract instructions from a meaningful 3-dimensional geometric relationship as a latent space. The signature profile is extracted from latent space which comprised five unique interpretations from Gabo’s Linear Construction.

Building from our combination of geometric logic with neural networking, we proposed a 900 square foot high-density foam bespoke suspended ceiling scape generated from an indicative of ruled surface sculpture convolutional neural networking and robotic fabrication. For the full-scale signature prototype mockup of our ceiling scape proposal, we used a large-scale hot wire end effector articulated by a six-axis robotic arm of IRB-4600 industrial robot by ABB Robotics.