Computational design and robotic fabrication of dry-stacked non-standard spanning limestone assemblies

Natural stone is an abundant, low-carbon material; however, quarrying and masonry processes generate substantial unusable waste, including rubble, offcuts, and irregular fragments. While interest in repurposing these byproducts is growing, their non-standard geometries pose major challenges for direct architectural integration, typically requiring further processing into aggregate or concrete. This study introduces a computational workflow to assemble spanning structures from irregular limestone waste through selective robotic modification. The methodology combines 3D scanning, heuristic optimization algorithms, and robotic milling to identify optimal stone pairings and minimize required machining. A generative algorithm evaluates a digitized library of 50 limestone fragments against a predefined structural geometry, determining which contact surfaces require planar modification and pin hole reinforcement as passive registration guides and shear keys. The workflow was validated through a three-legged arch prototype comprising 18 stones, demonstrating that non-standard quarry waste can be transformed into structurally coherent assemblies through computationally-guided selective intervention rather than full standardization. This approach preserves material morphology while enabling geometric control sufficient for spanning architectural applications.