Risk Avoidance Decision-Making for Intelligent Vehicles Based on Spatio-Temporal Risk Prediction

In order to improve the collision risk assessment method of intelligent vehicles in complex traffic scenarios and avoid collision risk in real time, this paper proposes a risk avoidance decision-making method for intelligent vehicles based on spatio-temporal risk prediction. Firstly, the risk measurement of spatio-temporal coupling is used as the evaluation index, which supervises the lateral and longitudinal collision risk of vehicles, and monitors the change of driving risk index in real time. Then, based on convolutional neural networks-bidirectional long short-term memory networks (CNN-BiLSTM), the spatio-temporal correlation of risk index is analyzed, and the potential collision risk is predicted to avoid driving risk in advance. The experimental results show that the proposed method can accurately evaluate the lateral and longitudinal risk. Compared with CNN and CNN-LSTM methods, the mean square error of the proposed method in different prediction horizon is reduced by 58.9% and 50.3<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> on average. Results show that the proposed method can effectively predict the potential collision risk.