Shared Control Strategies for Automated Vehicles

The advent of Automated Vehicles (AVs) holds promise as an efficient and safer alternative to manual driving. In addition to reducing human-caused accidents, it is also a matter of driving more efficiently with regard to the energy consumption, traffic
flow, and driver workload. However, these technologies are not yet mature enough for massive deployment in commercial vehicles, as assigning a passive role to humans in the driving task raises technical, social and legal issues. A technical limitation is that AVs cannot handle all driving situations, while a social problem is that humans like driving and have so far proven to be better drivers than machines. In this respect, the current approach in both the automotive industry and research community is not to replace the driver completely, but to let the driver and the automated vehicle cooperate within the traded control scheme (i.e., only one is charge of the driving task for a certain period). In this context, if the automation is responsible for the driving task, the driver is assigned the role of supervisor. Humans, however, have proven to be poor supervisors, as they tend to abuse automation, overtrust it, and are therefore unprepared to properly take back control when automation demands it. Furthermore, even attentive drivers are prone to lose situational awareness and may even become drowsy after a period of inactivity. As a result, any level of automated driving that partially or fully removes the human from the driver’s role presents a complex challenge that is still being explored. Yet, accidents happen all the time, and new solutions are needed to improve road safety. In this context, the shared-control approach offers the potential to improve driving safety and performance by keeping the driver involved in the control loop while taking advantage of recent advances in automated driving technology, such as perception systems, advanced control methods, and in-vehicle actuators. This control scheme allows the driver and the automation system to become a well-coordinated team, continuously working together at the tactical and control levels of the driving task. Therefore, more advanced and cooperative driver assistance systems are being sought than those currently available in commercial vehicles. AVs would then be the supervisors that the driver needs, rather than the other way around. Given these premises, the aim of this Ph.D. Thesis is to address both theoretical and practical aspects of shared-control in automated vehicles. First, a comprehensive review of the current state-of-the-art is performed to give an overview of the concepts and applications of shared-control that researchers have been working on over the last two decades. A practical approach is then taken by developing a steering shared-controller based on optimal control, which can assist the driver with different levels of haptic authority (intensity of torque) while maintaining performance and
stability in all cases. This controller and its associated decision-system (Arbitration Module) will be integrated into the general framework of automated driving and validated in a driver-in-the-loop simulator platform. In a final stage, two real-world scenarios are used to demonstrate the effectiveness of the controller. One is used to support a distracted driver, and the other is used to implement a safety function that enables overtaking maneuvers on roads with oncoming traffic. Both systems were evaluated with real participants using objective and subjective assessment methods. The conclusion of this dissertation is that shared-control is a promising approach for future automated driving features that can improve road safety in the short term while allowing people to continue to enjoy driving.