Our long-term goal is to build, test and validate an architecture for an agent that can support multiple goals in a dynamic environment of cooperative mobile agents. Initial tasks for our teams include surveillance and reconnaissance, search and destroy, pursuit, and evasion. A team of robots would be expected to perform these tasks with minimal supervision. Key components of this architecture were identified to be negotiation, strategic planning, execution and tasking control, execution monitoring, and recovery from failure. The challenge is to not only have several robots working together but to have them understand the effects of their actions on common team goals.
One challenge is that an agent may be working toward multiple, possibly conflicting, goals. Thus, the agent must be constantly evaluating its commitment to actions, or tasks, that contribute to the satisfaction of these goals. The imprecision of any action or sensory input has to be taken into account, and its contribution toward the satisfaction of current goals or plans assessed. In addition, the user must be kept informed of the progress of the team toward its goals. The user does not want to be actively involved in robot control, but must be able to intervene when necessary. Thus, monitoring must both ensure robust autonomous operation and provide the user with a window into the operation of the team.