Driving Intention Recognition of Human Drivers in Mixed Traffic Flow

The driving intention recognition (DIR) of human drivers in mixed traffic flow is a key issue in the intelligent transportation system, which plays an important role in accurate trajectory prediction and reasonable decision planning. To address this problem, we propose a DIR framework based on a deep neural network (DNN) which integrates the interactive information between the target vehicle and surrounding vehicles, road information, and vehicle state. The actual road NGSIM (Next Generation SIMulation) dataset is applied to verify our method. Compared with the widely used methods based on SVM (Support Vector Machines) and LSTM (Long Short-Term Memory), the proposed method is superior in precision, recall, F1 score, and accuracy. Its accuracy can reach 0.8888 and F1 score can reach more than 0.8555, which shows a good effect on DIR.