The Safety of Autonomy: A Systematic Approach

Abstract

Autonomy does not subvert existing safety processes, but they must be supplemented with methods that address autonomy's challenges, especially where perception and decision-making tasks are implemented with machine learning. We present an approach to address the safety of autonomous systems, building on and complementing established safety engineering methods. Traditionally, safety-related systems, like aircraft, cars, and factory robots, have operated under human control or supervision. With autonomous systems (ASs), the role of the human is lessened-perhaps just to the extent of initiating autonomous operation:An AS can operate independently of human control.ASs have existed for some time, for example, in rail, including the Docklands Light Railway in London. Although safety-critical, such systems operate within well-defined and, to an extent, controlled environments; for example, there are physical controls on human access to the tracks, and traffic movement is controlled through a signaling system. The introduction of such ASs has been successful, and they have good safety records.By contrast, emerging ASs, such as autonomous vehicles (AVs) on the roads or collaborative robots (called cobots) in factories, operate in significantly more challenging