Cognitive Assemblages: Spatial Generation Through Wave Function Collapse and Reinforcement Learning

This research explores the integration of AI in an iterative decision process for the open-ended procedural generation of architectural spaces. Leveraging on state-of-the-art Deep Reinforcement Learning techniques, an Artificial Neural Network (ANN) is trained to perform local decisions selecting tiles in a Wave Function Collapse (WFC) algorithm, assembling discrete elements that build up a complex spatial organization, pursuing selected spatial qualities at the architectural scale.