The effectiveness of route guidance and advanced control strategies in a corridor is a function of the time-dependent interaction between freeway and arterial flows in response to traffic conditions and information available to drivers. Although demand diversion is an essential element of optimal corridor management, no on-line demand predictor has considered explicitly the effect of control on the demand to be predicted. As a result, existing control strategies are mostly empirical and can address only known or anticipated peak-period loads. A new approach to modelling on-line demand diversion in freeway corridors is directly applicable to dynamic control and vehicle guidance. This new method treats freeway demand as a utility-maximizing, decision-making unit with ramp diversion behaviour dependent primarily on the travel delay caused by congestion, as modified by on-line information and traffic control strategies. The method combines behavioural modelling with an extended Kalman filter to identify the diversion model parameters recursively using the most recent prediction error in real time. Because diversion prediction does not require upstream flow information, system instrumentation requirements are substantially avoided. The method was used for on-line prediction of demand and diversion at freeway entrance ramps of the I-35w corridor in the Minneapolis-St. Paul metropolitan area. Test results from 5-min prediction on several weekdays indicate the entrance ramp diversion proportion is predicted with 90 to 95% accuracy.
Abstract