Human executive control of autonomous systems: A conceptual framework

Abstract

Autonomous systems are expected to be capable of performing many core functions with little or no human intervention. Many of the other functions required for overall success, however, still involve human supervision, manual control, or direct performance by a human. This paper presents the conceptual framework known as operator role theory, and specifically the concept of executive control (rather than supervisory control), in the context of autonomous systems. An autonomous ground transportation system (self-driving car) is used as an example. The functions associated with vehicle control and monitoring are expected to be under executive control. In conventionally driven vehicles these would be performed under manual or supervisory control. The fact that these functions will be performed under human executive control is the reason for calling the system autonomous. Such vehicles appear to be capable of driving the programmed route, controlling speed and lateral position, while avoiding roadway obstacles and collisions. They can execute normal driving maneuvers, and monitor vehicle health and the status of the trip against the plan, both spatially and temporally. A human executive controller will be capable of enabling and disabling these capabilities as desired, at least in some vehicles. Operator role theory provides a conceptual framework that will help guide understanding of the systems engineering issues, especially associated with function allocation, in designing and developing autonomous systems.