Emergency collision avoidance strategy based on the hierarchical predictive risk assessment method

In high-speed emergency collision avoidance scenarios, drivers may instinctively react with improper actions, leading to vehicle instability and loss of control. Aiming at this problem, this paper proposed an emergency collision avoidance strategy for advanced driver assistance systems (ADAS) based on a hierarchical predictive risk assessment method. The proposed method first uses a NAR neural network to predict the steering angle sequence over a specific future time. These prediction results are then used as the input for a discrete vehicle dynamics model to predict vehicle states. Once the predicted vehicle states exceed the safety boundaries, ADAS immediately intervenes and takes over the driver’s operation. The model predictive controller, which considers both vehicle stability and tracking performance executes the emergency collision avoidance maneuver. Extensive simulations and experiments conducted on the driving simulator demonstrate that the proposed strategy effectively avoids vehicle instability caused by the driver’s overreaction. Moreover, in comparison with existing studies, the proposed strategy exhibits superior adaptability to varying levels of scenario urgency.