Abstract: In order to sufficiently satisfy various travel demands and in the meantime take into account the development of urban economy and different travel modes, transportation management department has to optimize the travel mode structure according to the long-term contribution of all modes. Based on state-dependent utility function and logit model, this study builds a dynamic programming model which incorporates mode shares as the state variable as well as previous supply to reflect the influence of state inertial. MCMC algorithm from hierarchical Bayes method is introduced to estimate the weights of factors in the logit model. Then a four-share structure by the given shares of dynamic ridesharing and a transition probability matrix given the previous supply are constructed respectively. The simulation experiment displays the relationship of optimal policy and total revenue, the relationship of DR with previous supply and total revenue, the difference of mode structure between the beginning and end of period and, influence of total revenue on mode supply probability.

Key words: travel model, dynamic ridesharing, state-dependent utility, dynamic programming model