Automation technology and sharing economy have brought changes to transportation, such as the appearance of shared autonomous vehicles (SAVs). This paper proposes a mode choice model combining latent class model (LCM) and discrete choice model (DCM), to analyze choice behavior for SAVs and identify factors which influence the preference for SAVs. Stated preference survey was conducted to obtain data of four aspects, containing individual specific attributes, usual trip characteristics, individual behavior attitudes and alternative specific attributes. Considering the first three aspects of data as manifest variables, respondents are classified to four classes by LCM. The utility is formulated for these four classes and calibrated by multinomial logit (MNL) and mixed logit (MIXL) model. Estimation results show that the proposed latent class approach performs better than traditional MNL model in explanation ability, and significant influencing factors for each class are different, such as waiting time, travel time, number of travelers, usual travel modes, factors considered for mode choice, adopted SAV penetration rate. Results imply that the preference for SAVs differs across classes, and the preference for various modes differs across modes. The proposed approach can be used to analyze the effect of SAVs on travel behavior in the future.

CICTP