Prediction of Driver Deceleration Intent in Vehicle Platoon System

Abstract

This paper proposed the unscented Kalman filter (UKF) in which the state equations are defined by trend, car-following and artificial neural network models to estimate the headway and velocity of platooning vehicles. Based on the estimates, the authors then attempt to predict the driver deceleration intent using a conventional car-following model. The UKF integrated with the car-following model is the best among the other models, but it still needs some improvement to maximize the prediction accuracy.