Calibration of Microscopic Traffic Flow Models Considering all Parameters Simultaneously

This study proposes a methodology to calibrate microscopic traffic flow simulation models. The proposed methodology has the capability to calibrate simultaneously all the calibration parameters as well as demand patterns for any network topology. These parameters include global and local parameters as well as driver behavior and vehicle performance parameters; all based on multiple performance measures, such as link counts and speeds. Demand patterns are included in the calibration framework in terms of turning volumes. A Simultaneous Perturbation Stochastic Approximation (SPSA) algorithm is proposed to search for the vector of the model‟s parameters that minimizes the difference between actual and simulated network states. Previous studies proposed similar methodologies; however, only a small number of calibration parameters were considered, and none of the demand values. Moreover, an extensive and a priori process was used in order to choose the subset of parameters with the most potential impact. In the proposed methodology, the simultaneous consideration of all model parameters and multiple performance measures enables the determination of better estimates at a lower cost in terms of a user‟s effort. Issues associated with convergence and stability are reduced because the effects of changing parameters are taken into consideration to adjust them slightly and simultaneously. The simultaneous adjustment of all parameters results in a small number of evaluations of the objective function. The experimental results illustrate the effectiveness and validity of this proposed methodology. Three networks were calibrated with excellent results. The first network was an arterial network with link counts and speeds used as performance measurements for calibration. The second network included a combination of freeway ramps and arterials, with link counts used as performance measurements. Considering simultaneously arterials and freeways is a significant challenge because the two models are different and their parameters are calibrated at the same time. This represents a higher number of parameters, which increases the complexity of the optimization problem. A proper solution from all feasible solutions becomes harder to find. The third network was an arterial network, with time-dependent link counts and speed used as performance measurements. The same set of calibration parameters was used in all experiments. All calibration parameters were constrained within reasonable boundaries. Hence, the design and implementation of the proposed methodology enables the calibration of generalized micro-simulation traffic flow simulation models