Proportion-Based and Tendency-Based Bus Trajectory Prediction Models

Abstract

At the terminal station, the path headway bias of a trip is linearly correlated to the travel time of the path when it is beyond a critical range, which marks the tendency between successive trips. A tendency-based model for trajectory prediction was proposed in which the path headway bias was assumed to be proportionally amplified at downstream stops with an increase in path travel time, when buses fall into the amplifying tendency, or remain the same, when buses fall into the maintaining tendency. The boundary of the tendency was calibrated by the relative performance between the proportion-based model and the history-based model. The calibrated tendency-based model succeeds in avoiding the amplification of nondirectional path headway bias in the proportion-based model by keeping buses in the maintaining tendency and promoting the amplification of directional path headway bias by allowing buses to fall into the amplifying tendency.