Study of real-time traffic assignment: A rolling horizon approach

The need for a methodology to provide time-dependent estimates of traffic flows and associated performance measures in networks arises in several problem areas, including Advanced Traffic Management systems (ATMS) and Advanced Traveler Information systems (ATIS). Four different approaches, including mathematical formulations, optimal control theory, variational inequalities, and simulation-based approach, have been applied in developing Dynamic Traffic Assignment (DTA). However, these approaches focus on off-line dynamic traffic assignment problem. In order to achieve real-time control and operations, these models need to be extended to address real-time DTA issues. In this research, a real-time traffic assignment framework based on micro-simulation dynamic traffic assignment, utilized a Rolling Horizon approach, is proposed. The model addresses the problem where an ATIS/ATMS controller is assumed to have OD desires up to the current time interval, and short-term and medium-term forecasts of future OD desires. The core of the procedure is DYNASMART, a simulation-assignment model, and the model is used to replace link-performance and exit functions. Numerical experiments on a test network and a real network are conducted to investigate time-dependent flow patterns and real-time dynamics under a wide variety of assumptions on assignment rules.