Model development and its evaluation of predicting preceding car’s deceleration intent based on state estimation of vehicle platoon

Abstract

This paper proposes mathematical models to predict preceding car's deceleration intent and evaluates its effect on the following car driver in terms of the mitigation of risk of rear-end collision. Assuming a four-vehicle convoy, an unscented Kalman filter (UKF) integrated with a conventional car-following model attempt to estimate the headway and velocity of the 1st to 3rd vehicles in the platoon. The car following model used in this research is the conventional Gazis-Herman-Rothery model (GHR model). Then, another GHR model is used to predict the deceleration intent of the 3rd vehicle, which is expected to occur 1.5 seconds later based on the estimates of headways and velocities. In addition, this study proposes an interface to inform the predicted intention to the following vehicle by using a windshield display (WSD). Numerical analyses showed that the UKF integrated with the GHR model succeeded in estimating the vehicle platooning states and predicting the deceleration intent accurately. However, the prediction does not always shows the perfect match with the actual observation due to the uncertainty of driver's behavior. Although it is not completely accurate, it is a quite useful information to mitigate the collision risk so that we decided to design the interface by the WSD to inform the predicted deceleration intent to the following car driver. A driving simulator experiment was carried out to present the 3rd vehicle deceleration intent expected to occur 1.5 seconds later to the 4th ego car. It was found that the maximum deceleration rate of the 4th vehicle was suppressed and the value of the Time-To-Collision was increased not only in an ordinal deceleration case but also in a situation where the collision risk became significant.