An empirical delay model for application in unsignalized intersections in dynamic traffic assignment
MetadataShow full item record
Up until recently, unsignalized nodes have been either ignored or inadequately represented in Dynamic Traffic Assignment (DTA) models. This is due to the difficult nature of incorporating internal node conflicts into dynamic flow models. It was thought or assumed that these nodes had little impact on overall model results, but evidence from testing in Visual Interactive System for Transportation Algorithms (VISTA), a DTA model, reveals that may not be the case. This paper explores recent attempts at characterizing stop sign effects within DTA flow models. From previous studies, it has been found that incorporating these unsignalized and priority movements internal to the flow model requires large amounts of computational power, are challenging to make efficient, and lead to a multiple or infinite solution space. Based on these findings, a deterministic approach is both impractical and likely impossible in the existing framework of the Cell Transmission (CTM) and Link Transmission (LTM) models commonly used in DTA. Thus, a method of utilizing empirical relationships based on information readily available in these models may be a more acceptable approach. Microsimulation is much more suitable for modeling these types of interactions and is capable of producing results near to reality. For this reason, microsimulation was chosen as a viable method for developing empirical relationships of such complex interactions to then be used as inputs into the macroscopic flow models of DTA. This paper presents a model developed to calculate delays expected by vehicles at stop approaches based on information that can be taken from a dynamic flow model such as CTM and LTM models. This model is validated by video data recorded and analyzed for accuracy. Potential uses and probable implementations of the model are explored to appropriately incorporate unsignalized and priority movements into existing flow models.