The Frank-Wolfe algorithm is a common static traffic assignment (STA) solution used in travel demand models. Frank-Wolfe algorithms are capable of finding efficient solutions at a large scale due to their efficient computational nature. However, STA algorithms are generally criticized for: 1) their failure to address complex traffic flow phenomena, such as the formation and propagation of queues affecting corridor travel time; 2) only indirectly accounting for the impacts of corridor-level traffic control systems; and 3) the assumption that the least-cost path selection is based on a uniform cost function which applies to all users (irrespective of trip purpose, demographics, employment type, etc.). Dynamic traffic assignment (DTA) models are designed to resolve the problems identified above, but they require much greater computational resources, calibration efforts, and preparation of high-resolution data.
In this research, the team will implement the Frank-Wolfe algorithm and DTA with the same travel demand model input data to generate two sets of travel paths. The team will comparatively assess the energy use and emissions impacts of these two distinctly different traffic assignment algorithms and their resulting travel paths sets by applying MOVES-Matrix (to generate exact same output with MOVES).