Abstract: Granular materials (such as sand, gravel, and ballast particles) are widely used in geotechnical engineering, and understanding the mechanisms that influence their physical and mechanical properties is crucial for engineering practice. Numerous studies have shown that the macroscopic mechanical behavior of granular materials is closely related to geometric morphology of individual particles. However, existing studies mainly focus on single shape parameters and are often limited to monodisperse distributions. To address this, this study first establishes a comprehensive quantification framework for multilevel-polydisperse shape distributions based on the shape quantifications of realistic geotechnical particle assemblies. Subsequently, this study systematically develops methods to independently control polydisperse distributions of particle shape at each morphological level based on computational geometry theory. Finally, the effectiveness of the proposed particle generation methods is validated through discrete element simulations of isotropic compression tests. The simulation results indicate that the polydisperse distribution characteristics significantly affect the initial packing density and mean coordination number of the samples, regardless of the morphological level. This study provides a basis for systematically investigating the effects of multilevel-polydisperse shape distributions on the physical and mechanical properties of granular materials.