Data-driven design framework for AI-aided interactive architectural panels

Abstract

This thesis explores a workflow to develop an AI-aided, data-driven design framework aimed to achieve interactive panels on walls to improve user engagement. The study aims to achieve wall tiling tessellations to become interactive paneling systems, capable of responding physically to human presence. Using image and motion capture; these panels translate body movement into coordinated kinetic gestures, bridging a form of spatial communication between users and environments. Within the domains of computational design, artificial intelligence and interactive architecture, this research proposes a medium in which AI is housed in physical spaces. While researches of AI whithin the domain of architecture have typically been working on tools to help generate forms or optimize designs behind digital interfaces, this thesis explores its potential as a more physical layer embedded within architecture itself, allowing walls to sense, process and react to human activity in real-time. The thesis uses design-led research method to explore the interaction potentials within the defined design process and prototyping. This exploration bases upon four stages: (1) Case-study prototyping to explore the interaction system and needs, (2) Integration of AI to the interactive system (3) Simulation and re-creation of the system to generate different designs and experiment visual potentials, (4) design and fabrication of data-driven panels, which combines all knowledge gathered in the study. Moreover, The first stage focuses on the development of a data-driven panels' prototype using 36 display units inspired by Rozin's Weave Mirror. Using depth cameras and Arduino based controls enabled the prototype to interact and reflect the user's movement. The aim of the first stage is to investigate the necessary components and workflows needed for this project and how motion and perception can be represented using physical materials. Supporting the research with Gestalt psychology provides a conceptual base for how forms emerge through arrangement and proximity. In the second stage, artificial intelligence (AI) is introduced to aid the system's interactivity. Through Machine learning models, the panel is trained to recognize gestures, body positions, and voice inputs, each will trigger different responses and actions. These include expressive visual outputs or directional movement, transforming the panel into an interactive interface that evolves its behavior based on what it learns from the users interacting with it. The third stage centers on simulation and system creation. Digital tools are utilized to experiment with design variations, such as the number, shape and placement of display units. The fourth stage informs the construction and design of a second, more advanced prototype that enhances structural integrity, sensory responsiveness, and real-time feedback. Beyond the technical framework, the research in this thesis also reflects on questions of perception and engagement. The system operates between finding the balance between abstraction and recognition, prompting viewers to interpret meaning through motion. Design strategies rooted in Gestalt psychology to ensure that the interface remains perceptually understandable.