To study the advance distance of guide signs of expressway exit, we divide driving behavior in expressway exit into five processes according to the characteristics of the driving psychology principle and drivers' response to signs. A model of guide signs of expressway exit is built to calculate the advance distance of guide signs that meets safety and comfort requirements. The proposed model is established on the basis of the insertion gap travel distance model and isokinetic deviation sine curve model of vehicle lane changing distance. Through the analysis of the setting mode of guide signs, we calculate a reasonable distance for guide signs under different road conditions and different driver characteristics. The proposed model is developed with the UC-win/road and simulated by the Forum8 driving simulator. Simulation results show that only the distance of the advance signs of a two-way four-lane expressway exit is close to the specified value of Road Traffic Signs and Markings, which is reasonable; while for expressway with six lanes and above, the specified value is small.

Choice-making in diverging in an interchange is affected by multiple external factors. Along with an analysis of the characteristics of drivers' choice-making in expressway interchanges, this study settles many other fundamental variables, including the selection tree, utility function, speed of vehicles on the main line, accelerated speed, hourly traffic volume of the ramp, time headway in the deceleration lane, and type of vehicle, etc. The two-level nested-logit probability model for drivers accepting ramp diverging and drivers' diverging choice behaviors modes is established. Based on the radar tracking data of part of diverging areas of interchanges in Guanghe expressway, the model parameters are estimated by the phased estimation method. According to t-test result, the influence degree of the characteristic variables is judged and the model is optimized. Results suggest that the choice behaviors of diverging at interchange is affected by multiple factors comprehensively, the two-layer nested logit probability model is featured by a higher prediction accuracy.