|Source||CMES: Computer Modeling in Engineering & Sciences, Vol. 92, No. 5, pp. 477-491, 2013|
|Download||Full length paper in PDF format. Size =1,300,336 bytes|
|Keywords||route guidance, dynamic traffic assignment, model predictive control, network flow optimization, non-analytical iterative algorithm.|
Route selections for vehicles can be equivalent to determine the optimized operation processes for vehicles which intertwine with each other. This paper attempts to utilize the whole methodology of model predictive control to engender rational routes for vehicles, which involves three important parts, i.e. simulation prediction, rolling optimization and feedback adjustment. The route decisions are implemented over the rolling prediction horizon taking the real-time feedback information and the future intertwined operation processes into account. The driving behaviors and route selection speculations of drivers and even traffic propagation models are on-line identified and adapted for the simulation prediction in next prediction horizon. The mesoscopic traffic model is utilized for the simulation prediction so as to achieve both computing efficiency and prediction accuracy, where the partial link density in front of the vehicle rather than the density of total link is utilized to calculate the vehicle propagation velocity. The path traveling time is accumulated in a way related to the departure time and the operation process of a vehicle. The system architecture is composed of two parts. One is to simulate the true traffic system with stochastic behaviors such as speed fluctuations and inclinations to obey or disobey navigation commands, and the other one is the simulation prediction, rolling optimization and feedback adjustment system. In this way, the case study of medium traffic network shows that the simulation prediction-based rolling-horizon feedback implementation can prevent possible congestion in advance. It provides an engineering solution to the real-time closed-loop predictionbased traffic navigation.