Improved Car-Following Strategy Based on Merging Behavior Prediction of Adjacent Vehicle From Naturalistic Driving Data

As the fundamental control strategy of intelligent vehicles, car-following control directly affects vehicle performance. In practical driving, drivers usually predict the behavior of vehicles in the adjacent lane before modulating the driving strategy of the host vehicle. Therefore, an adaptive cruise control (ACC) system should be equipped with the practical ability to predict the following target in advance to improve the safety and acceptability of the intelligent control strategy. In this paper, a car-following strategy based on merging prediction of adjacent vehicles is developed from the results of naturalistic on-road experiments. Based on analysis of merging behavior parameters, the Fisher discriminant method is employed to establish a merging behavior prediction model of adjacent vehicles. Then, the desired spacing car-following model is ameliorated by the proposed merging prediction model. The simulation results of the proposed car-following strategy with different cut-in scenes indicate that the prediction model could forecast two kinds of merging behavior 2 s in advance, and the prediction accuracy rate reaches 88% and 90%, respectively. The improved car-following model could allow for smoother vehicle manipulation, thus enhancing safety and ride comfort. The results provide a reference for improving intelligent vehicle control algorithms and enhancing the acceptability of intelligent systems.