Optimizing the vulnerable units in the traffic network is the premise to ensure the stable operation of the road network. Therefore, the improved evaluation index which can comprehensively reflect the vulnerability of road network from many aspects is selected, the relationship between the vulnerability index and the number of road sections and traffic volume is analyzed, and the bilevel programming model is used to describe the discrete traffic network design considering vulnerability. The upper model takes the minimization of road network vulnerability as the optimization goal, and the lower model is the user equilibrium distribution model, By designing genetic algorithm and combining Frank Wolfe algorithm. The example results show that when the investment amount is 3 million yuan, compared with the scheme taking the vulnerability of a single road section as the optimization objective, the optimization scheme proposed from the perspective of the whole road network can reduce the vulnerability of 13 road sections to the ideal value, while the former is only 7. At this time, the scheme can reduce the vulnerability of the road network by 38.5%, which is 9.1% more than the former; It is also found that the greater the OD demand, the more construction funds need to be invested to reduce the road network vulnerability to the ideal value.