With the volume, torsional stiffness and transmission efficiency taken as the optimization objectives, a multi-objective optimization model of the RV reducer was constructed. To improve the design efficiency and save the computational cost, a Kriging surrogate model with a partial torsional stiffness was established based on the NXOpen C++ and the Abaqus Python secondary development technology. To solve the multi-objective mixed-integer nonlinear-programming problem, an MP-NSGA-II (mixed population-NSGA-II) algorithm was proposed, and the coding scheme for discrete variables was improved. An integrated RV reducer design software was developed by using PySide2, and the coupling relationship between optimization objectives was analyzed. The structural parameters selected by the entropy method were compared with BAJ-25E, and the effectiveness of this method was verified.