The sensitivity of empirically derived regression models of car-following performance is examined. Data were collected from coupled two-car units (a lead and a following car) in light traffic with programmed lead-car manoeuvres and in dense expressway traffic. Intervehicular spacing and velocities and accelerations for each of the cars were obtained. Models of changes of these variables with the operating condition and with the introduction of time delays in the variables were investigated using a computer simulation with empirical data. The principal findings were: (1) overall, car-following is quite linear, (2) the goodness of fit of any model is influenced by stream velocity and headway, lead-car velocity pattern, driver differences and time delays. These results were discussed in terms of the generality of car-following models with respect to the driving population, extent of coupling and the implications of lags in car-following performance for system ability.
Abstract