The objective of this paper is threefold. First, the feasibility of modeling a large-scale network at a microscopic level of detail is presented. Second, the unique data collection challenges that are involved in constructing and calibrating a large-scale network microscopically are described. Third, the unique opportunities and applications from the use of a microscopic as opposed to a macroscopic simulation tool are described. The possibility and feasibility of modeling a large-scale network using a microscopic simulation model is demonstrated. The requirements of a validated microscopic model for large-scale modeling are: ( a) the model must be capable of modeling origin-destination demand tables, ( b) the model must be capable of modeling dynamic traffic routing, and ( c) the model must be capable of modeling the dynamic interaction of freeway/arterial facilities. The data collection and coding exercise for microscopic models is more intensive than for macroscopic models. The calibration exercise for a microscopic model to a large-scale network, although feasible, is by no means an easy task and does require expert assistance. The Salt Lake metropolitan region study has demonstrated that the data collection, coding, and calibration exercise is approximately a 4-person-year exercise. Model execution times during peak periods are still quite high (from 2 to 17 times the simulation time depending on the number of vehicles) for the PC platform (Pentium 200 with 64 megabytes of random-access memory). Consequently, tools that can extract portions of the large-scale network can allow the modeler to conduct various types of sensitivity analyses within a more realistic time frame.
Using Trellis software to enhance high-quality large-scale network data collection in the field
