The work aims to optimize the air bag of the first generation of airborne vehicle and design the air bag for the safe landing of the second generation of airborne vehicle. Based on the existing air bag and combined with the finite element simulation and experiment design, the support vector regression (SVR) model with the air bag parameters as design variables and the impact acceleration peak value for vehicle landing as objective function was established. Genetic algorithm was used to optimize the parameters of the existing air bag, including its width, height, exhaust vent area and burst pressure of exhaust vent. The optimized air bag reduced the impact acceleration peak of the second generation of airborne vehicle by 73.4% and made the deceleration process of landing smoother. Parameters of the current air bag have been optimized and improved successfully to get the air bag suitable for the second generation of airborne vehicle. In addition, lots of air bag design schemes have been provided for the decision-maker.