I am a ph.d. student in Transportation Engineering program. My research topic is regarding day-to-day dynamics of the traffic assignment. What we need is to construct a rational learning process to model traveler's day-to-day route choice behavior. The limiting behavior is to reach a stochastic user equilibrium state, that is, no user can improve his perceived route travel time by unarily changing his route choice. At the same time, the system will reach a stationary route/link flow pattern. The first approach is a Markov assumption of user's route choice, which is the user's route choice on today is only dependent on the actual route travel time of yesterday. Then we can derive the invariant distribution of the route/link flow pattern.
My question is can we use Bayesian learning to model the traveler's learning behavior such that it will also converge to an equilibrium state, or an stationary distribution? If we can, how can we construct it? Thanks a lot, Kevin . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
