Non-Stationary Car-Following Dynamics: Driving Simulator Experiments

Abstract

Using a car-driving simulator created based on the open source engine---TORCS---we investigate human behavior in car-driving within the car-following setup. Five subjects with different skill of driving were involved in these experiments. Here we present the results of experiments comparing two particular versions of car-following setup; within one the lead car speed was fixed, within the other it changed in a certain irregular way. The latter setup was implemented for three different time scales of speed change representing "fast," "intermediate," and "slow" speed variations. In the case of "fast" speed variations a subject just cannot drive the car such that its speed and the lead car speed vary in time synchronically. In the case of "slow" speed variations it is an easy task. The histograms of the headway, velocity, acceleration, jerk, and the time derivative of pedal position (the main control parameter) constructed based on the collected data are compared. Previously we have demonstrated that driver actions should be categorized as human intermittent control. Based on the presented results we draw the conclusion that this type of driver actions is not sensitive to the uncertainty in the lead car motion.