Designing Human-Centered Autonomous Agents

Abstract

Human-centered automation (HCA) maximizes the goals of humans by supporting a full range of interactions between humans and autonomous systems. The key goal of this research is to minimize the necessity for human interaction, but maximize the capability to interact. Within HCA we define adjustable autonomy as the ability of autonomous systems to operate with dynamically varying levels of independence, intelligence and control. HCA encompasses adjustable autonomy and interfaces it with Human-Computer Interaction (HCI).

The motivations for human-centered automation are many. They include technical issues such as an inability to automate certain aspects of a task because of its complexity or because of an uncertain or changing operating environment. They also include non-technical issues such as a desire to allow for human intervention even when full autonomy is possible. These latter motivations may include safety, training, maintenance or calibration.

The benefits of human-centered automation include the ability to partially automate a system when full automation is not possible. Other benefits are lower costs because difficult-to-automate parts of the system can be left to humans and increased operator acceptance of an autonomous system.

Early work in human-centered autonomous systems has been conducted at NASA Johnson Space Center [Bonasso et al., 1997, Kortenkamp et al., 2000], at NASA Ames Research Center [Dorais et al., 1998], at the Honeywell Technology Center [Musliner and Krebsbach, 1999] and the University of Texas [Barber et al., 1999].