Analysis of interaction mechanisms between regular vehicles and autonomous vehicles for road safety purposes

In the context of an increasing number of surveys and campaigns pointing towards the immediate introduction of Connected and Autonomous Vehicles (CAVs) in the ordinary traffic, research might provide its contribution highlighting the benefits and the strength of these new vehicles, as well as the weaknesses and the potential issues deriving from their implementation. The aim of this research is to give a contribution in this field, by investigating the safe- ty-related aspects due to the introduction of CAVs in traffic. Since road safety is a wide area of investigation, especially if connected to the emerging technologies, it is necessary to focus on specific aspects and deeply investigate them. Considering that the available crash dataset in presence of CAVs are limited, current research highly relies on simulations of future scenarios, trying to capture the impact that CAVs can have on safety. In particular, traffic simulations can reproduce wide road networks by investigating the microscopic interactions between vehicles. One peculiar disadvantage of this ap- proach is that it strongly depends on the input parameters chosen. In fact, depending on the selected parameters, simulation outputs can drastically vary, sometimes lead- ing to unrealistic scenarios. While traffic simulations can simulate the interactions (and then, conflicts) between vehicles, they cannot explain specific individual and in- teracting driving behaviours in all time instants. This aspect can be explored relying on driving simulations, which can help in investi- gating specific driving behaviours in the traffic stream. Results from the simulations of given scenarios can then pave the ground for accurately calibrating traffic simula- tion parameters. This procedure can be particularly useful when CAVs should be sim- ulated, in absence of real on-road tests. This advantage provided by the driving simulator is the core of this study. In fact, this work uses driving simulation to study safety issues related to partially automated ve- hicles (PAVs) that still require human interventions and actions, and for this reason can lead to high risky situations. In fact, despite PAVs can help drivers in easy driving tasks, they require that the driver will eventually take control of the vehicle. This transition from automation to human intervention represents a blind spot for safety: if the driver is distracted, he cannot manage to take-over and then potential crashes or dangerous situations can occur. Considering the above reported background, driving simulation scenarios were de- signed and tested in this study. In detail, the main aspects investigated in this re- search are: • the effect of Adaptive Cruise Control (ACC), to understand the influence that this technology for advanced driving (a typical example of ADAS) has on the kinematic parameters of the vehicle (speed, acceleration) and on its position (lateral position and distance from the lead vehicle); • The effect of secondary tasks to be executed by the driver, to understand the responses and the readiness to take over by the eventually distracted driver; • the effect of repeating the same driving tasks, in order to test the potential in- fluence of familiarity with both the route and the driving tasks on the driving behaviour. A sample of 37 drivers, aged between 21 and 34 years old was recruited to test the simulation scenarios. The tests were run after collecting questionnaires and acknowl- edging the testers on their duties during the experiments. The scenario implemented in the driving simulator reproduced the geometric features of a tangent section belonging to one of the most crash-prone rural roads in the area of the Metropolitan City of Bari (MCB), namely the SP106 (a two-way two-lane rural provincial road). Data about this road section were available thanks to another re- search project related to the same area (MCB, Italy), which revealed the particular dangerousness of two-way two-lane rural roads, among the other road types. In the same research project, safety assessments were conducted, by considering also fu- ture scenarios involving the gradual market penetration of CAVs. Results from the driving simulation tests highlighted that, on average: • the ACC led to raise the average values of acceleration and distance from the lead vehicle and a reduction in the average value of speed and deceleration; • different typologies of secondary tasks led to different observed driving be- haviours and, in particular, visual distraction secondary tasks appeared to have greater negative impacts on the driving parameters; • drivers’ familiarity with ACC system and with the route was associated to an increase in the average deceleration value, a decrease in distance from the lead vehicle and a reduction of variability in lateral position. In conclusion, these results provide a contribution to understand potential issues re- lated to the introduction of CAVs with partial driving automation, which still require human intervention. Results can be useful for researchers since they could be used to calibrate traffic simulation parameters related to CAVs.