查看文章

The introduction of autonomous vehicles (AVs) introduces ethical issues when an inevitable collision scenario is detected with multiple outcomes/targets, eg deciding between steering to collide into an immovable rigid wall (IRW) or a group of pedestrians. In such cases, the decision making of a human driven vehicle is replaced by the decision algorithms on-board the AV. Therefore, the AV must be programmed to make ethical decisions, with this introducing the need for an ethical decision maker (EDM). As the AV collision targets can vary in terms of form and structure (ie an IRM or a pedestrian), a common cost unit for the collision injury severity levels is needed. Hence, the aim of this paper is to develop a mathematical framework using fuzzy logic that allows different collision targets to be compared onboard the AV in terms of the injury severity levels. Based on the pre-determined collision injury severity levels, an ethical decision can then be made. For example, if a utilitarian approach of Jeremy Bentham was to be programmed within the EDM, this would result in the AV steering into the collision target that minimises utility loss in terms of the severity of the road traffic injuries and/or number of fatalities.