Equilibrium models, which have for many years been the basis of travel forecasting in the form of static assignment models, are becoming more and more common in the area of dynamic traffic assignment (DTA) modeling. A DTAmodel is any model that assigns a time-varying origin-destination demand to a transport network in a time-varying way (e.g., expressed as time-varying path flows). This study is specifically concerned with DTA models thatare based on an underlying traffic model that satisfies the basic laws oftraffic flow theory as expressed in the well-known fundamental relationship between traffic flow, speed and density (often called the fundamental diagram of traffic). This mainly includes fluid-type traffic models (e.g. hydrodynamic) and traffic micro-simulation models. These types of traffic models can be combined with an assignment (or routing) algorithm in an iterative solution approach which converges to approximate dynamic user-equilibrium conditions. It should be noted that due to the inherent complexity introduced with this type of traffic model, these assignment algorithms arenecessarily heuristic in nature, though they are often inspired from well-known solution algorithms for the static assignment problem. The paper builds upon earlier work on an equilibrium assignment algorithm defined in the space of splitting rates (turning movement volumes by origin or destination), rather than path flows. It should be noted that splitting rate models have a structure that makes them well-suited for extension to en-route dynamic traffic assignment. The algorithm exploits the properties of the fundamental traffic flow relationship in computing a new set of splitting rates on each iteration. The splitting-rate computation explicitly considers the duration of the assignment time interval in conjunction with measures of link performance. The algorithm is used in conjunction with an efficient traffic simulation model, and is tested for the first time on real-world networks of significant size. The primary measures of performance are the number of iterations required to achieve dynamic equilibrium for a given scenario, and the proximity of the solution to true dynamic equilibrium conditions. For the covering abstract see ITRD E145999
Abstract