휴먼팩터를 고려한 보행자-자율차 상호작용 분석기법 개발

초록

This study proposes a methodology to quantify the relationship between autonomous vehicles (AVs) passengers’ psychological anxiety and traffic safety during interactions between AVs and pedestrians in mixed urban traffic. Electroencephalogram (EEG) signals collected from AV passengers are used to derive an Anxiety & Nervousness Index (ANI), representing their level of anxiety and nervousness, and its relationship with the Post-Encroachment Time (PET), a surrogate safety measure indicating pedestrian–vehicle conflict risk, is analyzed. Based on California AV crash reports and previous studies, three interaction scenarios—jaywalking, pedestrian island crossing, and mid-block crossing—are constructed, and scenario-specific PET and ANI values are computed in real time. Nonlinear regression models are fitted to the ANI–PET relationship for each scenario. The results show that, for the pedestrian island crossing scenario, the ANI–PET relationship is best described by a Gaussian function, whereas jaywalking and mid-block crossing scenarios are best captured by a double-exponential function. In all scenarios, ANI decreases nonlinearly with increasing PET, exhibiting a pattern of rapid drop, gradual relaxation, and saturation. Using an ANI threshold of 1.5 to indicate an anxious state, PET thresholds of 3.4 s, 3.0 s, and 2.8 s are obtained for jaywalking, pedestrian island, and mid-block crossing scenarios, respectively, below which AV passengers exhibit high anxiety levels. These findings suggest that, when AV passengers have no direct control authority over the vehicle, larger PET values than those reported in previous studies are required to secure psychological safety. The derived ANI–PET functions and thresholds can be used to define safe-entry conditions for different Operational Design Domains (ODDs), to develop AV performance evaluation metrics, and to support policies and educational materials aimed at enhancing pedestrian safety.

키워드