Automatically Generating Empirical Speed-flow Traffic Parameters from Archived Sensor Data

Traffic parameters used in the prediction models, which is traditionally based on assumptions, is one of the error sources for prediction. With the availability of traffic data in nowadays, the traffic data is popularly considered to apply for the validation of demand models. It is a crucial step for convincing the model prediction results. The flow-density model, which provides relationships between traffic flow variables, is a well-established approach for traffic prediction. Robust implementations of long- range planning and microsimulation models require calibration and validation of facility and time-dependent parameters (e.g. free flow speed, capacity, wave speed, critical density) which are sensitive to infrastructure, weather and other external factors. Archived data from freeway sensors provide a large sample from which to calibrate these parameters. In this paper, a set of automated traffic state identification tool is developed and applied to historical data to automatically determine the traffic phases. Once the traffic phase is known, we then calibrate the flow-density parameters with the fundamental diagram. Using this tool, with the identified traffic state for many days’ data, the traffic parameters for free flow state and congestion state can be calibrated automatically in a fundamental diagram. As a case study, the calibrated traffic parameters in a dynamic traffic model (Cell Transmission Model) CTM are presented. The main findings of this paper is that an automated parameter calibration method is applicable for practical use, and this method provides a convincing result.