| Summary: | Dynamic traffic assignment is widely recognised as being more useful to evaluate traffic management measures than is static counterpart, as it allows us to analyse how congestion forms and dissipates in time-varying conditions. In this thesis, both deterministic and stochastic dynamic assignments are modelled with a proper link performance function, and solved with efficient solution algorithms so that they give rise to high quality solutions. A deterministic dynamic assignment is formulated in the form of variational inequality and solved by a route-based solution algorithm which intrinsically respects correct flow propagation. Similarly a stochastic dynamic assignment is formulated in the form of variational inequality, but solved with a link-based algorithm with an explanation on how to maintain correct flow propagation in this solution approach. In particular, both solution algorithms are developed in a way that we can find optimal solutions efficiently without direct evaluation of an objective function, based on the interpolation method. In both dynamic assignment techniques, the deterministic queuing model is adopted as the basis of the link performance function. This model is suitable to describe the relationship between inflows, outflows, and travel costs for a link in time-varying conditions because it respects all requirements for dynamic traffic modelling such as traffic conservation, the FIFO discipline, correct flow propagation, and causality. Finally, application of both dynamic assignment techniques to several test networks, including a medium-size network with 24 nodes and 76 links, shows that a proper way of associating costs with flows in discrete time is crucial to the calculation of plausible dynamic assignments.
|
|---|
