Evaluating the Driving Safety of Heavy Trucks of Third-Party Logistics Companies in the Absence of Crash Information

Evaluating the driving safety of heavy trucks has always been a bottleneck problem for the third-party logistics company because of the lack of sufficient crash-related information. In this study, the naturalistic driving data of heavy trucks is used to extract risky driving indicators (RDIs). The risk level of each driver is evaluated using the three RDIs (acceleration propensity, overspeed propensity, and shifting gear propensity indicators). Finally, K-means clustering is employed to classify the risk level of each driver. Two road sections with a length of 1,050 and 220 km are selected to compare whether the risky behavior of the same driver is different under the influence of mileage. The results show that the risk level of the same driver on road sections with different lengths is significantly inconsistent, and high-risk behaviors will be more concentrated on a few drivers. The conclusions can help the 3PL company to evaluate the driving safety of heavy trucks without crash data support.