Multivariate Regression for Estimating Driving Behavior Parameters in Work Zone Simulation to Replicate Field CapacitiesAuthor/Presenter: Edara, Praveen; Chatterjee, Indrajit
Traffic simulation tools are being increasingly adopted by DOTs for various traffic analysis studies, including assessing work zone traffic impacts. This is due, in part, to their ability to model individual vehicle and driver behavior at a highly detailed level to assess the traffic performance. In order to accurately use the simulation models for traffic analysis of work zones, it is necessary to calibrate the models to match the field conditions (such as lane capacity and queue lengths) by adjusting the driving behavior parameters. Unfortunately, there is very little guidance on choosing the driving behavior parameters in simulation models that will replicate the actual field conditions in work zones. Empirical studies have shown that the roadway capacities at work zones are not only lower than the capacities under normal operating conditions but also vary across all states in the US. This disparity means that a unique driving behavior parameter set cannot be used by all states; instead the parameter values should be chosen so as to reproduce the state-specific capacity values. The default truck characteristics inside the VISSIM simulation model are different from the average characteristics of trucks on US freeways. The use of default truck characteristics that are not representative of the US conditions results in incorrect calibration and therefore erratic conclusions. Based on a review of national studies on truck characteristics (length, weight to power ratio, etc) we recommend US specific values for use in VISSIM. The recommended values are generic and are applicable to facilities other than work zones. In this study, we develop multivariate regression models to express the relationship between the critical driving behavior parameters of the VISSIM model, work zone lane configuration, truck percentages, and work zone capacity. The estimated statistical models are capable of generating a range of parameters’ values that produce a wide range of capacities used by state DOTs in the US. Model validation is also conducted using field data from work zone sites in Ohio. The validation results showed that the model recommended values are capable of producing accurate estimates of capacity and queue lengths.