To overcome the shortcoming that driving manipulation qualification can’t be embodied by current driver models, the automatic optimization ability of genetic algorithms is adopted to summarize drivers’ self-learning features and driving experience, off-line optimize fuzzy PID scale factor and quantization factor by following running error minimum and physical ability-to-pay minimum principle, and simulate the driving training processes from a green hand to a skilled driver. Then a driver-vehicle closed loop system simulation model including direction fuzzy PID optimized by genetic algorithms, speed fuzzy integrated control driver model and entire vehicle riding dynamics model is established, which simulates and analyzes typical modes, such as longitudinal speed one-way variation, lateral double lane and big curvature test road. The simulation results show that the model can well describe drivers’ longitudinal acceleration/deceleration manipulating characteristics and lateral desired track following steering riding characteristics. And when it’s compared with traditional PID and fuzzy PID, it has better longitudinal acceleration/deceleration manipulating characteristics and lateral desired track following steering riding characteristics.