The automated wingman - Using JACK intelligent agents for unmanned autonomous vehicles

摘要

The Belief, Desires, and Intention (BDI) agent model is used in operations analysis and computer generated forces. It has been shown that the BDI approach to agent programming brings the particular advantage of using terms and concepts that are familiar to all. At present the challenge is to bring this a step forward with appropriate tools and a control environment that enable pilots to operate an uninhabited airborne vehicle (UAV) during mission by means of tactical directives. The BDI model provides a suitable framework for capturing tactical behaviours in the air operation domain. The representation of these behaviours in terms of BDI plans is readily understandable by domain experts. However the current development process requires an experienced BDI developer to work with a domain expert in order to develop an appropriate set of agent behaviours. We would like to remove the BDI developer from the loop - this paper represents a first step in the achievement of that objective. The idea is that for controlling a UAV as an 'autonomous wingman', the pilot instructs the UAV by forming and commanding tactical plans at a suitable level of abstraction. The autonomous wingman would process these plans in the context of common-sense background behaviour. Technically this control environment is a simplified programming environment, where the pilot is presented with a collection of programming 'primitives' of high-level tactical behaviours and world-events based on the platform's sensor system. By combining primitives into plans, the pilot can build up a sensible instruction sequence for the UAV to accomplish particular tasks in ways appropriate for the situation at hand during mission.