Based on finite element method and control volume method, a finite element model of equipment and its airbag cushion system were established and verified experimentally. The simulation of an impact of a few tenths of a second duration typically required tens of hours of CPU time. As a result, the optimization of the design based on a nonlinear model was very difficult by traditional iterative approach method. In order to overcome this problem with the design optimization of airbag cushion system for airdropping equipment, surrogate models were employed instead of the complex finite element model based on Extended Latin Hypercube method and Radial Basis Function. The height of airbag, the width of airbag and the area of vent hole were chosen as design variables. Then Pareto optimal solution sets based on response surfaces were obtained by multi-objective Genetic Algorithm. After optimization, the maximum acceleration reduces 19%, while the maximum attitude angle reduces 1%.The results show that the cushion performance and stability of airbags cushion system are obviously improved through optimization.