Car-following models look microscopically at the traffic stream and consider reaction of the individual driver to his surroundings. These models are developed to help understand the behaviour of traffic, especially during congested conditions, and are an essential component of micro-simulation models of traffic and transport networks. The latter are increasingly being used in research and practice to design and evaluate transport schemes and policy strategies. This paper describes and general approach to calibrate and validate car-following models using commonly available loop detector data. The method was designed for calibration on an open stretch motorway section rather than in a closed ring structure, making it more applicable to real-world calibration applications. The calibration was formulated as an optimization problem which seeks to minimize the discrepancy between the observed and the modelled traffic flow and average speed. The minimization problem is solved through an interactive procedure which loops through a feasible range of parameter values and identifies the set of values that has the minimum objective function. The method was demonstrated through the calibration of a newly developed car-following model designed to model motorway traffic. The model divides the traffic conditions into different states: free-flow, close-following, flow breakdown and traffic recovery from a breakdown. Different model parameters are used to describe the different states. In the demonstration, two sets of model parameters representing free-flow and traffic breakdown states were calibrated. The results suggested that the calibration solution algorithm had converged and the calibrated parameters were of reasonable values.(A) Reprinted with permission from Elsevier. For the covering abstract see ITRD E134766.
Abstract