Automated detect and avoid: Autonomy and ethics

摘要

In the aeronautical, automotive and nautical domain, automation of collision avoidance maneuvering is being addressed. In the automotive domain, the potential occurrence of so-called dilemma situations imposes yet unresolved challenges to the design of the automation. In the aeronautical domain, systems enabling (pilot-supervised) automated selection and execution of collision avoidance maneuvers have been certified. To prevent the need for a collision avoidance maneuver, aircraft have to remain Well Clear. For today's Detect and Avoid systems, the pilot has to choose and initiate the Well Clear maneuver. In this paper, a roadmap for a stepwise increase in automation of Detect and Avoid systems is discussed and the associated challenges are identified. It is argued that, similar to the automotive domain, beyond a certain level of automation, the ethical question under what conditions pilot judgment can be automated must be answered, and that this may well determine the upper limit to the level of automation that is acceptable.