Equilibrium Method for Origination Destination Matrix Estimation Exploited to Urban Traffic Simulation Calibration

One of methods used to find better light signalization plans for urban regions is a traffic flow numerical simulation. In this paper we present solution of two main issues the user of traffic flow simulation tool encounters while simulating traffic continuously registered in ITS by counting detectors, namely: (a) origination-destination routes setting, (b) calibration of the driver model. We consider two methods for origination-destination matrix estimation. The first is classical Spiess gradient method and the second one is the equilibrium method of Casseta which is suitable to congested scenarios. Resulting origination destination matrices from both methods are input to the traffic simulator to generate traffic between nodes in the city. The traffic generated in the simulator in this manner is used to calibrate the average driver model, to achieve as good recovery of vehicle counts on detectors as possible. As a simulation tool we use an open source ArsNumerica Execution Environment running on top of SUMO simulator. The driver model that is used is stochastic Krauss model. Calibration is done for one of the most congested areas in the city of Wrocław in the morning peak.