The Air Force Human Resources Laboratory (AFHRL) has established a program to design, develop, and validate an expert model of pilot decision-making in Air Combat Maneuvering (ACM) to be used in the training of fighter aircraft pilots. The intent of this program is to create a computer-based simulation which can encapsulate the expertise of combat pilots in ACM strategy, tactics, and offensive and defensive decision-making. The resulting expert system is to be incorporated into a flight simulation package to support the training of these ACM skills to student combat pilots. The development of the ACM Expert System is based on the latest advancements in the technology of Artificial Neural Systems (ANS).

To effectively train student combat pilots in ACM skills, it is desirable to move beyond the textbook, allowing the student to interact with a simulated adversary aircraft via computer. Unfortunately, it is difficult to capture ACM expertise in computer software which will provide the student with realistic and reasonable adversary behavior. Most existing systems use "pre-canned" profiles or simple trajectory generators. The more advanced adversary simulators use rule-based expert systems to represent and recall pilot expertise and create a more reactive system. However, traditional expert systems suffer from major inadequacies which have limited their success. By using an ANS approach to this problem, actual human ACM performance data is being used to "train" an expert system in ACM decision-making skills. This system is capable of simulating human ACM performance by learning to associate the recognition of a tactical situation with the selection of the proper course of action. The objective of this paper is to describe current efforts to apply ANS technology to the training of ACM decision-making skills.