The advent of Cooperative, Connected, and Automated Mobility (CCAM) is already underway, yet its safety implications remain under-explored. Hard Braking (HB) events in automated shuttles provide crucial insights into safety performance, particularly when analysed across multiple naturalistic pilot sites. This study employed Negative Binomial (NB) Regression to examine HB event frequency across 10 real-world operational sites within the EU Horizon 2020 SHOW project [1] and also considered site-specific variations. Data were collected from 10 European pilot sites: Brno, Carabanchel, Graz, Karlsruhe, Klagenfurt, Les Mureaux, Linköping, Pörtschach, Tampere, and Trikala. These sites provided 1,796 daily observations including 4,820 HB events, with cutoff limits determined by the automated shuttle operators. The statistical modelling included NB regression due to significant overdispersion in the data, with the variance exceeding the mean by a factor of 20.15. The model incorporated average speed, acceleration variance, and pilot site categories as explanatory variables. A random intercept approach was used to account for site-specific variations, with Linköping serving as the reference category. To verify multicollinearity concerns, Variance Inflation Factors (VIF) were calculated, with the highest value being 2.333, ensuring robust model interpretation. Additionally, Marginal Effects to the Mean (MEM) were computed to quantify the impact of predictor variables on HB counts. Both average speed and acceleration variance were found to significantly increase HB occurrences (p < 0.001). A higher average speed led to a proportional rise in HB events, likely due to more abrupt braking requirements in response to traffic interactions or unexpected obstacles. The acceleration variance, which captures fluctuations in driving smoothness, also exhibited a strong correlation with HB events, indicating that higher variations in acceleration contribute to harsher braking events. The results emphasize the necessity for adaptive and context-aware automation strategies to mitigate HB events. Future research should incorporate additional factors, such as pedestrian interactions, intersection dynamics, and urban layout characteristics, to refine safety models further. The study’s findings provide a foundation for improving AV deployment strategies, guiding regulatory policies, and optimizing automated shuttle operations across diverse European urban environments.
| ID | pc591 |
| Presentation | |
| Full Text | |
| Tags |
