On applying machine learning to develop air combat simulation agents

Several approaches for utilizing machine learning technologies towards improving the capabilities of autonomous, simulation-based agents are described. For an autonomous agent to be robust, it must be able to plan its activities, react quickly to unforseen events, and execute planned or modified behaviors to achieve goals. Autonomous agents that exhibit appropriate behavior for simulated air combat, providing intelligent, realistic adversaries and cooperative allies, are under development. Building such agents is not trivial, and the techniques of machine learning hold great promise for extending the capabilites of hand-coded systems. The application of some of these techniques, past successes, and current research directions are described