Evaluation of car-following models through accurate experimental data

Car-following behaviour, fundamental element of the traffic microscopic analysis and simulation tools, has always need to face the data problem of accuracy and consistency. In the past, data were collected through complex mechanical and electric systems in controlled traffic flow conditions. Recently new techniques such as video and radar screening have highlighted clear limits in the spatial analysis approach for the first, and in the effort of synchronizing data in multiple car-following experiments for the second. The recent development of satellite positioning techniques (K DGPS) opens a new horizon in traffic engineering, allowing multiple car-following data collection with extreme accuracy, in real traffic flow conditions and for large time periods. The first research objective is, therefore, to develop a methodology for the collection and the multiple car-following data elaboration through K DGPS. The proposed methodology has been applied to gathered data: a series of experiments developed with four vehicles on urban and extra-urban roads has allowed the collection and the calibration of the speed and acceleration profiles, of the spatial and time distances among vehicles and the reaction times of the drivers in different flow conditions. Data collected have been used to make a comparative evaluation of two car-following models: the model developed by Gipps (1981) and the consequent changes proposed by Punzo (2001, 2003).