In order to establish the car-following fuzzy inference system (FIS) accurately，the fuzzy clustering
analysis method is adopted to divide input and output variables into various fuzzy sets，and the corresponding Gaussian membership function is determined． Then，the Takagi-Sugeno inference method is introduced to conductthe car-following fuzzy inference and the defuzzification． Finally，the fuzzy inference system and the Gippscar-following model are calibrated and evaluated by using NGSIM data． The results show that the improved method to establish the car-following fuzzy inference systemcan truly reflect the characteristics of the data itself and the driver’s psychological and physiological features，and that the established fuzzy inference system achieves a much higher simulation accuracy of error indexes in comparison with the Gipps model．