A non-linear model for signal timing optimization was created to study the problem of four phases signal timing optimization at city's single-point intersections.The minima of average delay and average stop times were regarded as the optimization objective function,the phase green time and the cycle time were regarded as the constraints and one of the Shanghai intersections was regarded as the research object.With the real-coded genetic algorithm in the Matlab simulation environment,the optimization problem was solved.By applying the method,the intersection cycle time changes from 145 s to 118 s,shortened by 19%,the average vehicle delay changes from 45 s/veh to 36 s/veh,shortened by 20%,and the average stops changes from 0.828 2 to 0.736 1,shortened by 11%.The results show that the signal timing plans drawn by using this method can reduce parking delay and stops effectively,thus it proves the usefulness of the model which is better than the existing control scheme and the Webster optimization algorithm.