Modeling Humans for Coordinating Distributed Human-Agent Teams in Space Operations

Current manned space operations consist of crew members performing predefined procedures assisted by flight controllers located in Mission Control at Johnson Space Center (JSC). For future operations such as space exploration, where missions are extended and communication delays may be a factor, we are investigating new concepts of operations. Specifically we are developing software agents to assist both crew and flight controllers in their duties. We have developed control agents for crew and vehicles system that reduce human workload by automating routine operations such as reconfiguration and fault management. We also have developed human support agents that aid in distributing operations. These agents relieve humans of vigilant data monitoring by automatically drawing human attention to important events when they occur. They allow flight controllers to perform some operations from their offices. And they increase crew autonomy by providing services such as event notification, procedure tracking, and schedule management that reduce the need for close tending by flight controllers. Taken together these humans and agents form an operational team. To coordinate effectively, team members need to be aware of each other. This includes knowing where team members are located, what activities they are performing, what roles they fulfill, and if they are healthy. This information is used both as a context for communication as well as a means of coordinating activity. Tracking human state can be difficult, however, because you often are reasoning with incomplete or incorrect knowledge. Humans can be difficult to sense because you must rely on sensing the effects of their actions and the environment they occupy and manipulate often is poorly instrumented. Humans also are difficult to model because they tend to do things in different ways over time and they can commit a variety of errors. We have developed the Distributed Collaboration and Interaction (DCI) System to assist humans and automated control agents working together in support of distributed space operations [1]. The DCI system includes Ariel agents that observe their human users’ location and activities for the purposes of team coordination. In this paper we describe these software agents. We discuss our approach for modeling the human in these software agents and how we use this model to support coordination among the human-agent team. We then describe our experience in using these agents with a life support application at JSC.