EXTENDED ABSTRACT: Stacking two-dimensional (2D) crystals is a unique way to induce novel physical properties and technologies. The interfacial friction is near zero as two crystal surfaces contacting incommensurately, which is termed as structural superlubricity (SSL). The synergistic effects of multiple factors such as material properties and interface structure on the friction have led to unclear regulation mechanisms for SSL, hampering the development of the fundamental research and the applications of SSL. We developed a high-throughput program based on first-principles methods to study the interfacial friction, achieving efficient acquisition of friction performance of van der Waals (vdW) homo/heterogeneous interfaces. Combining graphene, penta-graphene (PG), B2N4, and transition metal chalcogenides to construct homo/heterostructures, introducing incommensurate contacting through interface twisting and stretching, and analyzing their effectiveness in reducing interface friction. An explanatory machine learning (ML) model was constructed to describe friction based on high-precision computational data. A two-step ML model was employed to search the superlubricants. This project exerts the functions of modulating, constructing and platform provided by vdW stacking, and achieves rapid search of superlubricants via combining the high-throughput calculations and ML, reveals the important factors affecting friction performance, and attempts to establish a theoretical model to describe the interface friction.Keywords：structural superlubricity; van der Waals structures; machine learning; high-throughput calculations.