Evaluating the traffic impacts of work zones is vital for any transportation agency to plan and schedule work activity. Traffic impacts can be estimated by using microscopic simulation models. One challenge in using these software models is obtaining the desired work zone capacity values, which tend to vary from state to state. Thus, the default parameter values in the model that are suitable for normal traffic conditions are unsuitable for work zone conditions, let alone for conditions specific to particular states. Although a few studies have been conducted on parameter selection to obtain desired capacity values, none of them have provided a convenient look-up table (or chart) for the parameter values that will replicate field-observed capacities. Without such provision it has not been possible for state agencies to use any of the research recommendations. This study provides the practitioner a simple method for choosing appropriate values of driving behavior parameters in the VISSIM microsimulation model to match the desired field capacity for work zones operating in a typical early-merge system. The two most significant car-following parameters and one lane-changing parameter were selected and varied to obtain different work zone capacity values. CC1 is the desired time headway, CC2 is the longitudinal following threshold during a following process, and the safety distance reduction factor is representative of lane-changing aggressiveness. It has been verified that the recommended parameter values not only produce the desired capacities but also create traffic conditions consistent with traffic flow theory.

A microsimulation calibration methodology based on matching speed-flow graphs from field and simulation is presented. Evaluation and automation of matching speed-flow graphs are based on methods from pattern recognition. The methodology was applied to the US-101 freeway network in San Francisco, California, by using an evolutionary algorithm. The methodology is compared to traditional methods of calibration based on capacity and is shown to perform better. In addition, a small-scale test-network simulation model was developed to assist in calibration of large-scale simulation models. The performance of the test-network-based calibration is comparable to the US-101 simulation model.

No abstract