A Learning-Based Day-to-Day Dynamics Model for Frequency-Based Transit Assignment

Passengers’ travel experience on previous days will potentially influence their transit route choices on the next day. The day-to-day information is an important reference for decision-making of the transit assignment. This paper investigates the frequency-based transit assignment problem in a learning-based, day-to-day dynamics environment taking the learning and prediction behavior into consideration. The passengers’ route cost updating, as well as the flow updating, are explicitly considered and incorporated in the day-to-day learning process of the transit assignment model. With the proposed cost updating and flow updating process, the frequency-based optimization model of the day-to-day transit assignment problem is formulated. Afterwards, the model is solved with the method of successive averages. Finally, the numerical experiment validates the effectiveness of the proposed model and algorithm. It is revealed that the passengers’ flow evolution process will tend to be stable due to the day-to-day learning mechanism.