The calculation of path travel times is an essential component for the dynamic traffic assignment and equilibrium problems. This paper presents a simulation method for calculating actual path travel times for the traffic network with dynamic demands. The method is based on a path-based macroscopic simulation model of network traffic dynamics. There is no need to explicitly model intersection delays in this method. Discontinuity in the travel time caused by traffic light control can be captured by this method. It's flexible in terms that the model is not limited to a specific velocity-density relationship. Some numerical results for signalized and unsignalized networks are reported. / Singapore-MIT Alliance (SMA)

Urban travel demand forecasting and traffic assignment models are important tools in developing transportation plans for a metropolitan area. These tools provide forecasts of urban travel patterns under various transportation supply conditions. The predicted travel patterns then provide useful information in planning the transportation system. Traffic assignment is the assignment of origin-destination flows to transportation routes, based on factors that affect route choice. The urban travel demand models, developed in the mid 1950s, provided accurate and precise answers to the planning and policy issues being addressed at that time, which mainly revolved around expansion of the highway system to meet the rapidly growing travel demand. However, the urban transportation planning and analysis have undergone changes over the years, while the structure of the travel demand models has remained largely unchanged except for the introduction of disaggregate choice models beginning in the mid-1970s. Legislative and analytical requirements that exceed the capabilities of these models and methodologies have driven new technical approaches such as TRANSIMS. The Transportation Analysis and Simulation System, or TRANSIMS, is an integrated system of travel forecasting models designed to give transportation planners accurate, and complete information on traffic impacts, congestion, and pollution. It was developed by the Los Alamos National Laboratory to address new transportation and air quality forecasting procedures required by the Clean Air Act, the Intermodal Surface Transportation Efficiency Act, and other regulations. TRANSIMS includes six different modules: Population Synthesizer, Activity Generator, Route Planner, Microsimulator, Emissions Estimator, and Feedback. This package has been under development since 1994 and needs significant improvements within some of its modules. This dissertation enhances the interaction between the Route Planner and the Microsimulator modules to improve the dynamic traffic assignment process in TRANSIMS, and the Emissions Estimator module. The traditional trip assignment is static in nature. Static assignment models assume that traffic is in a steady-state, link volumes are time invariant, the time to traverse a link depends only on the number of vehicles on that link, and that the vehicle queues are stacked vertically and do not traverse to the upstream links in the network. Thus, a matrix of steady-state origin-destination (O-D) trip rates is assigned simultaneously to shortest paths from each origin to a destination. To address the static traffic assignment problems, dynamic traffic assignment models are proposed. In dynamic traffic assignment models, the demand is allowed to be time varying so that the number of vehicles passing through a link and the corresponding link travel times become time-dependent. In contrast with the static case, the dynamic traffic assignment problem is still relatively unexplored and a precise formulation is not clearly established. Most models in the literature do not present a solution algorithm and among the presented methods, most of them are not suitable for large-scale networks. Among the suggested solution methodologies that claim to be applicable to large-scale networks, very few methods have been actually tested on such large-scale networks. Furthermore, most of these models have stability and convergence problem. A solution methodology for computing dynamic user equilibria in large-scale transportation networks is presented in this dissertation. This method, which stems from the convex simplex method, routes one traveler at a time on the network and updates the link volumes and link travel times after each routing. Therefore, this method is dynamic in two aspects: it is time-dependent, and it routes travelers based on the most updated link travel times. To guarantee finite termination, an additional stopping criterion is adopted. The proposed model is implemented within TRANSIMS, the Transportation Analysis and Simulation System, and is applied to a large-scale network. The current user equilibrium computation in TRANSIMS involves simply an iterative process between the Route Planner and the MicroSimulator modules. In the first run, the Route Planner uses free-flow speeds on each link to estimate the travel time to find the shortest paths, which is not accurate because there exist other vehicles on the link and so, the speed is not simply equal to the free-flow speed. Therefore, some paths might not be the shortest paths due to congestion. The Microsimulator produces the new travel times based on accurate vehicle speeds. These travel times are fed back to the Route Planner, and the new routes are determined as the shortest paths for selected travelers. This procedure does not necessarily lead to a user equilibrium solution. The existing problems in this procedure are addressed in our proposed algorithm as follows. TRANSIMS routes one person at a time but does not update link travel times. Therefore, each traveler is routed regardless of other travelers on the network. The current stopping criterion is based only on visualization and the procedure might oscillate. Also, the current traffic assignment spends a huge amount of time by iterating frequently between the Route Planner and the Microsimulator. For example in the Portland study, 21 iterations between the Route Planner and the Microsimulator were performed that took 33:29 hours using three 500-MHZ CPUs (parallel processing). These difficulties are addressed by distributing travelers on the network in a better manner from the beginning in the Route Planner to avoid the frequent iterations between the Route Planner and the Microsimulator that are required to redistribute them. By updating the link travel times using a link performance function, a near-equilibrium is obtained only in one iteration. Travelers are distributed in the network with regard to other travelers in the first iteration; therefore, there is no need to redistribute them using the time-consuming iterative process. To avoid problems caused by link performance function usage, an iterative procedure between the current Route Planner and the Microsimulator is performed and a user equilibrium is found after a few iterations. Using an appropriate descent-based stopping criterion, the finite termination of the procedure is guaranteed. An illustration using real-data pertaining to the transportation network of Portland, Oregon, is presented along with comparative analyses. TRANSIMS framework contains a vehicle emissions module that estimates tailpipe emissions for light and heavy duty vehicles and evaporative emissions for light duty vehicles. It uses as inputs the emissions arrays obtained the Comprehensive Modal Emissions Model (CMEM). This dissertation describes and validates the framework of TRANSIMS for modeling vehicle emissions. Specifically, it identifies an error in the model calculations and enhances the emission modeling formulation. Furthermore, the dissertation compares the TRANSIMS emission estimates to on-road emission-measurements and other state-of-the-art emission models including the VT-Micro and CMEM models. / Ph. D.