Abstract: At present, the most common approach for vehicle aerodynamic drag reduction is trial and error method, which completely depended on the engineers' experience. Inevitably, this method is certain blindness and low efficiency. Moreover, it is quite difficult to parameterize the auto body in the aerodynamic optimization design. To solve these problems, a fast and effective parameterization method named FFD (free form deformation) was introduced into the vehicle aerodynamic optimization design. Firstly, taking the Ahmed model as the simulation object, sample space was constructed based on the orthogonal array experiment design, and FFD method was used to parameterize the model at each sample point. Secondly, the aerodynamic drag coefficient of each sample was confirmed by CFD simulation. Then, according to the sample space and computational results obtained by CFD, three kinds of approximate model were constructed, and the RBF model with highest credibility was selected to build the approximate model. Finally, the multi-island genetic algorithm was employed to get the optimal combination of the design variables. To validate the reliability of the final results, a new model was reconstructed according to the results of the optimization, and aerodynamic drag coefficient was computed by CFD again. The computational results show that the Ahmed model aerodynamic drag coefficient can be decreased by 51.96% after optimization, which confirms that FFD method deserves to be popularized in vehicle aerodynamic optimization design.