Abstract:
Microscopic traffic simulation models have been widely used to analyze the traffic operation
and management strategies on highways and urban streets primarily because the simulation is
less disruptive to traffic, less expressive, and faster than the field experimentation. In order to
provide accurate and meaningful results, simulation models must be calibrated and validated
before real simulation is conducted. However, calibration methods differ in algorithm, level of
accuracy, convergence time and level of effort. It is essential to choose a calibration method
particularly suitable for a specific traffic and roadway condition.
In this research, a new approach has been proposed for calibration of microscopic traffic
simulation model VISSIM. Wiedemann 99 car following model and lane changing model
parameters were simultaneously calibrated for freeways. Drone was used to capture real time
traffic data over a 2.6 km stretch segment of the Dhaka-Mymensingh Highway (N3) in
Bangladesh for the calibration and validation through image processing technique. New
mathematical equation, based on the fusion of microscopic and macroscopic traffic data, has
been derived dynamically from Leader-Follower pair of simulated vehicle trajectory data to
evaluate the measure of performance (MOP) between the observed and simulated traffic data.
Root mean square error (RMSE) between the measurement of space headway of 1191 instances
of observed lane change and space headway derived from the mathematical equation by
evaluating simulated vehicle trajectory data has been considered as the fitness function for
calibration, and sum of absolute error (SAE) for validation against average headway
measurement. Three optimization techniques namely, Genetic Algorithm (GA), Simultaneous
Perturbation Stochastic Approximation (SPSA) and Simulated Annealing (SA) were used to
examine their performances in terms of level of accuracy and computational time in calibrating
iii
the parameters. In the perspective of present study on non-lane based mixed traffic condition
with MOP derived from new mathematical equation based on micro and macro data, SA proved
to be efficient reducing RMSE to 93.90%, while GA reduced to 92.15% and SPSA 91.85%.
Though the microscopic simulation is getting increasingly common in traffic planning and
operational research, the most important drawback of traffic simulation model is that the
calibration and validation of such model can be very tedious. This invokes the necessity of a
graphical user interface (GUI) for ease in calibration and result interpretation. In this study a
user-friendly Optimization Program Control Interface (OPCI) has been demonstrated for auto
calibration that controls the entire process of calibration and generates the desired graphical
output for result interpretation and analysis.