Empirical Validation of Three Macroscopic Dynamic Network Loading Models

The macroscopic dynamic network loading (DNL) model is the essential component to perform dynamic traffic assignment (DTA). The popular DNL models widely used in literature of DTA are Point-Queue (PQ) and Linear Travel Time (LTT) models because they fulfil of all the required desirable properties and require lesser computational cost. The Divided Linear Travel Time (DLT) model is presented as a variant model to overcome theoretical deficiencies found in PQ and LTT models. Literature of DTA focused on theoretical properties of these models; however, empirical validation of these models with real traffic data is scarce. This study prime objective is to validate PQ, LTT and DLT models with collected data in terms of inflow and outflow profiles from simultaneous recorded videos at both ends of the link under various conditions of link volume. The obtained results indicated that models are not able to replicate observed travel time profiles as Root Mean Square Error (RMSE) values are quite higher in almost all analysed cases. Investigation of properties such as FIFO and heterogeneity in transport modes using a fastest lane experiment indicated that model predictions are better when extent of these measures were lower. It implies that incorporation of these notions in future models are means to further progress in this direction.