Interactive Vehicle Trajectory Prediction Considering Abnormal Emotion Based on SOR Cognitive Framework

The precise anticipation of vehicle trajectory is a fundamental technological requirement for vehicle active safety systems to work, especially in scenarios with strong interaction. The majority of existing algorithms for trajectory prediction rely on the driving behavior exhibited by drivers in a normal emotional state. The impact of abnormal emotions is not taken into account. It leads to low prediction accuracy of trajectory, resulting in false alarms. To address this issue, this paper built a CARLA-SUMO co-simulation platform and recruited 26 volunteers, who were induced to experience abnormal emotions of anger and fright for driving simulation experiments in scenarios with strong interaction. Based on Dynamic Bayesian Network (DBN), the key causal chains from anger, and fright to severe collision risk are deduced. We constructed the trajectory prediction scheme of Double-GCN-BiLSTM-Attention with reference to the stimulus-organism-response (SOR) cognitive framework. This model integrates emotional factors and enhances the precision of trajectory prediction. This study proves that the introduction of the SOR cognitive framework can significantly improve the accuracy of trajectory prediction. The research results can be promoted to car companies, and the feasibility of landing is strong, which is of great significance to improving driving safety and creating social and economic value.