Fly-by-agent: Controlling a pool of UAVs via a multi-agent system

This paper describes the multi-agent system used to control a package of four uninhabited air vehicles (UAVs). The system has recently been used in a series of test flights where the pilot of a fast jet controlled a team of four UAVs (one real, three simulated) carrying out a representative mission. The structure of the system is described and the re-organisation of the agents as the mission progresses is illustrated with an example. The paper concludes by describing the importance of whole system issues and the integration and test cycle for getting AI techniques working and accepted in an application.