Human-Agent Teaming for Effective Multirobot Management: Effects of Agent Transparency

The U.S. Army Research Laboratory is engaged in a multi-year program focusing on the human role in supervising autonomous vehicles. We discuss this research with regard to patterns of human/intelligent agent (IA) interrelationships, and explore the dynamics of these patterns in terms of supervising multiple autonomous vehicles. The first design pattern focuses on a human operator controlling multiple autonomous vehicles via a single IA. The second design pattern involves multiple intelligent systems including (a) human operator, (b) IA-asset manager, (c) IA-planning manager, (d) IA-mission monitor, and (e) multiple autonomous vehicles. Both scenarios require a single operator to control multiple heterogeneous autonomous vehicles, and yet the complexity of both the mission variables and the relations among the autonomous vehicles makes efficient operations by a single operator difficult at best. Key findings of two recent research programs are summarized with an emphasis on their implications for developing future systems with similar design patterns. Our conclusions stress the importance of operator situation awareness, not only of the immediate environment, but also of the IA's intent, reasoning and predicted outcomes.