Studying the safety impact of sharing different levels of connected and automated vehicles using simulation-based surrogate safety measures

Resum

CAV (connected and autonomous vehicles) are becoming a reality and are gradually but steadily infiltrating the markets. CAV have promised improving traffic safety and are anticipated to do away with mistakes made by human drivers. Accordingly, the number of traffic safety studies involving connected and autonomous vehicles (CAV) has recently increased. Because there is a lack of information about the real behaviour of CAV in mixed traffic flows, traffic simulation platforms are used to provide a reasonable approach for testing various scenarios and fleets. Various traffic microsimulation platforms with distinct traffic flow models (e.g. Aimsun, VISSIM, PARAMICS, and SUMO) have been used in the literature, where studies have reported that CAV may improve traffic safety, particularly in high sharing percentage scenarios. Nevertheless, the exist research was either limited for including a calibration of one or two levels of automation, or do not analyse and present the results in term of the effect of including each level of CAV. Moreover, the severity term was not clearly stated and discussed in these investigations. Further, a check of the sensitivity of the usual parameters used for CAV calibration on traffic safety has not been addressed before. This doctoral thesis aims to assess the impact of near-real-time introduction of CAV into the traffic flow with varying levels of automation (from Level 1 to Level 4) on traffic safety in terms of quantity and severity, taking into account the fact that Level 4 vehicles won't be introduced into the traffic right away. The thesis also aims to evaluate the safety impact of a proposed scenario for CAV introduction; operating the CAV on dedicated lanes. Lastly, the thesis endeavors to highlight the most influential factors of driving dynamics from a traffic safety perspective. The investigation began with the modelling of various CAV levels using Gipps' model calibration, followed by the simulation of nine mixed fleets of CAV levels on a simulated highway segment. Following that, the Surrogate Safety Assessment Model has been used for vehicle trajectory safety analysis. According to the findings, gradual penetration of CAV levels resulted in a progressive reduction in traffic conflicts. This reduction ranges from 18.9% when 5% of the vehicles on the traffic flow have high levels of automation (Level 3 and Level 4 vehicles) to 94.1% when all vehicles on the traffic flow are Level 4. Furthermore, human-driven vehicles and vehicles with low levels of automation (Level 1 and Level 2 vehicles) are more frequently involved in conflicts (as potential inductors of risky situations; as follower vehicles) than vehicles with high levels of automation (Level 3 and Level 4 vehicles). In fact, depending on the combination of different types of vehicles in the traffic flow, human-driven vehicles are involved in conflicts from 8% to 122% more than their fleet sharing percentage, whereas vehicles with high automation levels are involved in conflicts from 80% to 18% less than their fleet sharing percentage. Increasing interaction with CAV on roads reduces the severity of conflicts, especially for vehicles with high levels of automation (Level 3 and Level 4 vehicles). Level 4 vehicles operation result in conflicts with the lowest severity. Afterwards, in general, in relation to investigating the effect of using a dedication lane during the CAV introduction period, it is found that not using a dedicated lane for a penetration rate up to 55% (Level 3 and Level 4 vehicles) provides better safety outcomes than using a dedicated lane for light traffic condition, whereas, using a dedicated lane is better always in congestion conditions. Finally, by exploring the influence of driving parameters in calibration on traffic safety, the main key parameters that show significant influence on traffic conflicts are reaction time, clearance, maximum acceleration, normal deceleration, and sensitivity factor. Further, by exploring the influence of the interaction between each two of these key parameters, the results show that, a low maximum acceleration when combined with other parameters’ values, always generate the highest number of conflicts, whereas short reaction time combinations always produce the best traffic safety results. On one hand, this thesis confirms the theory and previous literature conclusions that indicate a safety gain due to CAV penetration. On the other hand, it offers a broader perspective and support for the implementation of CAV levels. Furthermore, this study sheds light on how many potential conflicts could arise as serious conflicts during the transition period from a fully manual vehicle operation scenario to a fully CAV operation scenario. As a result, this thesis broadens both manufacturers' and researchers' perspectives on CAV behaviour for future implementation.