Influencing factors of risky behavior in truck safety: A random parameter model incorporating trip-wise heterogeneity

Truck-related crashes cause significant economic losses and casualties, making perception and control of truck driving risk a critical task for the logistics industry. However, heterogeneity in each truck trip is ignored in current studies, hindering precise prediction of truck driving risk. In this research, we defined trip-wise driving behavior as the driving characteristics extracted from real-time trajectory during a single trip, and investigated its impact on truck driving risk considering heterogeneity. Multi-source data were collected and aggregated, including on-board device data recording long-term trajectory and conflict events from 4,672 trucks in China, accompanied by high-resolution traffic environment data collected at the same time. We extracted trajectories on the same route for trucks within the selected fleet, and illustrated the existence of heterogeneity in trip-wise driving behavior using the Kruskal-Wallis test. A random parameter logit model was employed to study the influencing factors on truck driving risk, considering trip-wise heterogeneity. Results indicated that the heterogeneity of each truck trip was mainly reflected in standard deviation of trip-wise speed and environmental conditions (e.g., traffic speed, time of day). The effect of higher standard deviation of trip-wise speed varies significantly across trips, decreasing risk in 73.7% trips and increasing risk in 26.3% trips; this variability was shown through the normal distribution of the estimated parameter. Furthermore, heterogeneity shows the complex factors influencing truck driving risk and reveals overlooked patterns in long-term and trip-wise driving behavior, highlighting the importance of combining long-term behavior pattern with trip-wise behaviors for better risk prediction.