BrickNet: A computational design framework for robust dataset creation for robotic fabrication and design automation

Generative AI models have gained considerable attention across various fields, demonstrating remarkable success in generating text and image data. This paper presents the first phase of a comprehensive three-step project focusing on the development of a data creation pipeline for robotic fabrication. In this phase, we propose a computational framework in Grasshopper3D for the parametric generation and structural analysis of non-standard brick wall designs. We validate the framework by comparing the performance of two physical simulation results of two engines, ABAQUS CAE and Nvidia PhysX, highlighting critical insights into structural stability of the walls without mortar. While this phase does not include AI-based generative design or robotic fabrication, it establishes a robust foundation for future research. The findings provide essential data structures and simulation protocols for subsequent deep learning model training and physical robotic assembly. Finally, the benefits and limitations of this simulation-driven approach are critically analysed, suggesting improvements and future avenues for integrating generative AI into robotic fabrication workflows.