A Multi-Agent System for Package Level Autonomy
Jeremy W. Baxter, Graham S. Horn · 2004
*† In this paper we describe how a multi -agent system can be used to allow the pilot of a fast jet to control a package of str ike UAVs on a time -critical targeting mission. We concentrate on the benefits of adopting a multi -agent approach and de scribe how the agents interact with the other parts of the mission system. Results from using test pilots to fly missions in a synthetic environment are briefly discussed. Previous trials on similar missions had required the pilot to control each UAV independently and feedback from the pilots had indicated that the workload associated with selecting and tasking each UAV to match the mission was too high. What was desired was a ‘pool’ approach whereby the pilot issued orders to a group of UAVs which then self -organiz ed to achieve the task s. As part of the most recent trial we therefore implemented a set of increasingly autonomous tasks, carri ed out by a multi -agent system, to allow the pilot to operate at a variable level of autonomy to reduce their overall workload .