Simultaneous Off-Line Demand and Supply Calibration of Dynamic Traffic Assignment Systems

The off-line calibration of Dynamic Traffic Assignment (DTA) systems using aggregate traffic sensor measurements has generally been achieved by focusing on system components independently, and is often restricted to fitting only vehicle counts data. This paper proposes a new, general framework for estimating all DTA system parameters simultaneously, while efficiently incorporating any available prior parameter estimates and a wide range of readily available sensor data (including time-varying counts and speeds from loop detectors). An efficient algorithm is developed to estimate both demand and supply model parameters. The algorithm combines two simulation optimization approaches to span the parameter space and refine the search for a global optimum through quadratic fitting and minimization. The resulting estimator, possessing very good convergence properties, is tested on a prototypical network with synthetic data through an extensive experimental design. Dynamic Network Assignment for the Management of Information to Travelers (DynaMIT), a simulation-based DTA system with traffic estimation, prediction and route guidance capabilities, is used as the test system. Extensive numerical results are used to illustrate the usefulness and robustness of the developed estimator, and draw preliminary conclusions regarding the value of the approach.