Calibration and validation of a micro-simulation model for operational performance evaluation of urban corridors

Abstract

The use of traffic analytical and simulation models has gained more recognition in recent years from the transportation and traffic engineering community as alternatives to conventional methods for traffic congestion alleviation. These models are employed for the design, analysis and management of the transport system. The models offer a cost-effective means of assessing the effect of congestion alleviation alternatives on the transport system performance before implementation. Traffic analytical models such as SIDRA are considered as individual intersection models. However, when SIDRA is employed for the analysis of urban corridors with successive intersections, the impact each intersection has on the performance of the other is not considered and therefore, the actual performance of each intersection on the corridor is not truly reflected. For performance analysis of intersections locally, SIDRA is presently used and performance improvement alternatives are warranted, irrespective of the nature of the road network. This then raises concerns on the performance analysis adequacy of such networks. To analyse urban corridors with successive intersections, simulation models such as VISSIM (a micro-simulation model) are employed because of their ability to capture and assess movement interactions of intersections within the network. Moreover, the applicability of the models for local traffic conditions must be investigated as none of the models were developed using South Africa’s traffic conditions. The objectives of the study were to evaluate the performance improvement variations at two signalised intersections that were upgraded, and then to calibrate both SIDRA and VISSIM for the local urban corridor with successive signalised intersections. The better performing model was then validated using a new dataset different from that used for calibration purposes. The model was further used to perform an overall network performance analysis. The performance improvement variation was established by comparing before and after performance data with model predicted performance data (based on delay and LOS) for morning and afternoon peak hours (07:00 am – 08:00 am and 16:30 pm – 17:30 pm, respectively). An urban corridor with four successive signalised intersections was calibrated against field-measured performance (with respect to delay, LOS and travel times). At the two upgraded intersections, significant variations between the models’ predicted performance improvement and the obtained performance after upgrades were observed. The model (SIDRA) showed overestimation and underestimation behaviour regarding performance improvement predictions. The two models’ (SIDRA and VISSIM) calibration results compared to field performance illustrated that the VISSIM model more accurately replicated field performance. The validation results of the calibrated VISSIM model well replicated the field performance showing a strong correlation between the two data sets. The validation results showed that for both delay and travel times the VISSIM model error results were lower than 12% for both MAPE and RMSE. The overall network performance showed that the corridor segment evaluated was operating at an undesirable level of service. The study concluded that isolated performance analysis of intersections might not present the actual performance of the intersection where an urban corridor consisting of successive intersections is concerned. Therefore, the VISSIM model can be additionally used for holistic performance analysis on local urban corridors. It is recommended that further research on the local applicability of the VISSIM model for performance analysis of larger urban networks and on urban corridors with mixed intersections be considered.