Application of Different Data Analytics for Evaluation of Heavy Vehicle Vulnerability in Cold-Region Rural Highways

Abstract

Understanding the impacts of different driving conditions on truck speed is critical to the development and maintenance of resilient highway freight transportation systems. This study attempts to evaluate the combined impact of road weather, travel lane, vehicle type, and truck payload conditions on drivers’ speed choice by modelling speed distributions as normal distributions. Two data analytics, a regression-based approach (RBA) and a central limit theorem (CLT)-based approach (CBA), are adapted to model context-specific speed distributions. The regression-based approach models population-level speed distributions by considering samples of individual speed data, whereas the CBA uses sampling distributions produced according to the CLT. A holistic approach is proposed to identify overall vehicle-specific collision risks imposed by different road-weather conditions, based on the speed distribution parameters estimated. Implications of the study results pertaining to the trucking industry are threefold. First, adapting different data analytics leads to different study results; yet, the CBA is recommended to model speed distributions. Second, truck speeds are significantly affected by the presence of adverse road-weather conditions, yet marginally varied under different loading conditions. Third, overall, tractor–trailer combinations (TTCs) entail high collision risks, particularly when transporting a freight load under adverse road-weather conditions. The study results would be useful to policymakers, particularly for effective speed management in extremely cold regions. Trucking companies may use the study results to identify the least risk-posing road-weather conditions to deploy safe freight transport operations. The resulting speed distribution models are also useful as input to calibrate traffic micro-simulation models.