A Car-Following Model Based on Trajectory Data for Connected and Automated Vehicles to Predict Trajectory of Human-Driven Vehicles

Connected and Automated Vehicles (CAV) have been rapidly developed, which, inevitably, renders that human-driven autonomous vehicles share the road. Thus, trajectory prediction is an important research topic, which helps each CAV to efficiently follow a Human-Driven Vehicle (HV). In wider scope, prediction, also, improve throughput of traffic flow enhance its stability. To realize Vehicles, car-following model, based on data, was established. Adding deep neural networks Attention mechanism, this paper established data-driven CNN-BiLSTM-Attention for CAV, predict trajectory, by referring modeling idea traditional model. The data in next-generation-simulation (NGSIM) datasets match characteristics were selected. addition, noise-reduction pre-processing performed, make it actual situation. Experiments, selecting optimal structure model method carried out. models, such as LSTM, GRU, CNN-BiLSTM, selected comparative analysis prediction. results show has smallest MAE MSE well largest R2. highest accuracy vehicle-trajectory can, effectively, provide theoretical basis vehicle-trajectory-based velocity guidance Vehicles. future, decisions mixed flow.