Take-Over Performance and Safety Analysis Under Different Scenarios and Secondary Tasks in Conditionally Automated Driving

Abstract

In conditionally automated driving, the driver is required to takeover control of a vehicle if a takeover request is issued due to possible system limitations. This study investigates the effect of roadway environments and secondary tasks on takeover performance and safety. The experiment was conducted in a real vehicle-based driving simulator. Participants experienced three different traffic scenarios, including a non-critical scenario and two critical scenarios. Manual driving, a 1-back cognitive secondary task and a letter game task were each tested in each scenario. Results indicate that different driving scenarios and secondary tasks impact takeover characteristics; in particular, they have a strong effect on takeover time and driver workload, which impacts takeover safety. Specifically, the steering reaction was generally slower than the braking reaction, indicating that lateral operation requires more cognitive and decision-making time. In extreme cases, braking operation alone was insufficient to ensure safety, and steering operation was required. When obstacles were difficult to detect, or when the driver was engaged in a visual secondary task, the steering reaction time increased significantly. This study provides data for takeover safety evaluations and best practices of conditionally automated driving. INDEX TERMS Automated driving, control transitions, takeover time, test scenarios, secondary tasks. I. INTRODUCTION Automated driving has developed rapidly in recent years. Following the introduction of automated driving systems by Google and Tesla, traditional automakers, such as BMW, Mercedes and Audi, have also released similar systems and carried out road tests [1]. However, there are still many problems that need to be solved for driver acceptance and driving safety [2]. According to a Google report, over 14 months of testing, its self-driving cars were 'out of automated status' 272 times, and there were 69 instances in which drivers chose to cancel the automated status [3]. This shows that the automated system still relies on human intervention. According to SAE's J3016 standard, most mass-produced automated vehicles are still under Level 3, i.e., conditionally automated driving [4]. This means that the system is only suitable for The associate editor coordinating the review of this manuscript and approving it for publication was Daxin Tian.

Employing Emerging Technologies to Develop and Evaluate In-Vehicle Intelligent Systems for Driver Support: Infotainment AR HUD Case Study

“…As the simulation progresses, the division of the driver's attention secondary to the external stimulus (i.e., text message and/or navigation message) creates a situation of a potential imminent collision. In a short interval after receiving the message, a rear collision scenario is unfolding; as the lead vehicle, every time is performing an abrupt braking or full-stop braking [50,59]. The latter is appearing on a seemingly random part of the simulation as a traffic jam is blocking the road ahead, in order to minimise the likelihood of the user memorising the events.…”

A Refined Non-Driving Activity Classification Using a Two-Stream Convolutional Neural Network

“…For instance, visual related activities tended to take longer reaction time than the auditory related activities [8]. To achieve high-quality takeover and safety enhancement [9], it is therefore crucial to precisely identify, distinguish and track the type of NDA that the driver is engaging with, then to evaluate the status and attention level or workload for the improvement of vehicle safety and operational efficiency. However, there is very limited literature focusing on that.…”

Intelligent Mobility in the City: The Influence of System and Context Factors on Drivers’ Takeover Willingness and Trust in Automated Vehicles

“…However, there is much less research on these contextual factors. The criticality of the situation and the scope of action, manipulated by the number of lanes or the available time for a driver to act, are context factors that influence driver's behavior in manual driving (Heesen et al, 2012;Stoll et al, 2020a;Stoll et al, 2020b) and takeover scenarios (Wu et al, 2019) as well as the acceptance of behavior choices of an automated vehicle (Stoll et al, 2021). With reducing the scope of action, the situation was perceived as more critical (Stoll et al, 2020b) and the acceptance of automated behavior increased (Stoll et al, 2021).…”

“…With reducing the scope of action, the situation was perceived as more critical (Stoll et al, 2020b) and the acceptance of automated behavior increased (Stoll et al, 2021). When the criticality of the situation was higher, takeover time was shorter, and takeovers were more critical and unsafe, which was reflected in a shorter time to collision to the front vehicle (Wu et al, 2019). Another context factor that was explored more extensively in the literature is the complexity of the situation, which is often associated with the traffic density in a scenario.…”