To solve the problems such as lack of advance guide signs of small spacing segment exits on urban expressway, this study proposed five kinds of advance sign setting schemes and conducted a driving simulation experiment. Nine types of evaluation indicators were extracted and analyzed by repeated ANOVA, and six indicators were selected which were significantly affected by different schemes. TOPSIS method combined with entropy weight coefficient was used to evaluate the effectiveness of the five schemes. Results showed that it had the best effect on the driving behavior when setting advance guide signs at 1 km, 0.5 km, and 0 km before the starting point of the deceleration lane. It was followed by the scheme with four advance guide signs according to China standard (GB 5768.2—2009), and the scheme with four advance guide signs according to Japan related standards. K-means cluster and LSD comparison analysis revealed that the effectiveness of setting one advance guide sign was poor. Since the scheme of three advance guide signs had the best effectiveness, it is recommended to be adopted. When road conditions are limited, at least two advance guide signs should be set.