A Model-to-Decision Approach for the Autonomous Vehicle (AV) Ethical Dilemma: AV Collision with a Barrier/Pedestrian(s)

摘要

The introduction of autonomous vehicles (AVs) is expected to reduce the number of road traffic accidents; however, it is expected that this will not be completely incidents free. In the event of AV incidents, ethical decisions will need to be made, e.g. a decision between colliding into an immovable rigid barrier or a group of pedestrians. Selecting the collision target of such an ethical problem involves the investigation of a model-to-decision (M2D) approach. Thus using a mathematical model of the collision scenarios to make decisions based on the severity of each outcome, e.g. minor or severe and potentially fatal injuries. In this paper, a novel nonlinear (bilinear) lumped mass-spring model of an AV colliding into an immovable rigid barrier has been developed. This model is used in conjunction with a developed pragmatic ethical decision maker (EDM) approach. This research investigates the use of such an approach for AVs (containing one or two occupants) in collision scenarios involving the choice between colliding into an immovable rigid wall or into one/ten pedestrian(s). Assuming the utility of those involved in each outcome is equal, a possible solution to the ethical problem is known as the utilitarian approach; this leads to the AV steering into the collision path of least fatalities. To gain a better understanding of the public acceptance of how an AV EDM should be programmed in the case of collisions, a survey was undertaken. The results suggest that participants of the survey support the idea of AVs steering into the collision path of least overall severity, even if the number of persons at risk is greater. The results from the M2D approach highlight the effectiveness of using such an approach on-board future AVs to minimise the severity of road traffic injuries and/or the number of fatalities.