Calibration of dynamic traffic assignment models with point-to-point traffic surveillance

Abstract

Accurate calibration of demand and supply simulators within a Dynamic Traffic Assignment (DTA) system is critical for the provision of consistent travel information and efficient traffic management. Emerging traffic surveillance devices such as Automatic Vehicle Identification (AVI) technology provide a rich source of disaggregate traffic data. This thesis presents a methodology for calibration of demand and supply model parameters using travel time measurements obtained from these emerging traffic sensing technologies. The calibration problem has been formulated in two different frameworks, viz. in a state-space framework and in a stochastic optimization framework. Three different algorithms are used for solving the calibration problem, a gradient approximation based path search method (SPSA), a random search meta-heuristic (GA) and a Monte-Carlo simulation based technique (Particle Filter). The methodology is first tested using a small synthetic study network to illustrate its effectiveness. Later the methodology is applied to a real traffic network in the Lower Westchester County region in New York to demonstrate its scalability.

(cont.) The estimation results are tested using a calibrated Microscopic Traffic Simulator (MITSIMLab). The results are compared to the base case of calibration using only the conventional point sensor data. The results indicate that the utilization of AVI data significantly improves the calibration accuracy.