An approach to intelligent decision making in computer generated forces

The need for artificial intelligence and intelligent decision making as its core is in the very essence of Computer Generated Forces (CGF) research. It is crucial in order to obtain a realistic representation of both the natural habitat of the battle zone (citizens, vehicles, etc.) and of the threat the force will encounter. Three levels of decision making exist and differ by the models suitable for each one: reactive, short-term and long-term. Decision of all three types should be handled by the CGF, as well as combinations and compositions of decisions. Many models exist for decision making, each with its advantages and its faults. We at the Computer Generated Forces development team at Battle Lab had examined those models and found that none of the existing models is completely suitable for our needs. We have therefore developed a new, innovative model which combines two basic artificial intelligence models-fuzzy controllers and state machines-to create a whole greater than the sum of its parts. This model can serve to solve all three types of decision making, and by it nature is suitable for decision composition. It is also built as an infrastructure, so that it can be used not only by the CGF, but by other applications in the simulation as well, where they require autonomous decision making or decision support for the human user. This model is currently in use in the CGF, solving both algorithmic problems such as collision avoidance and behavioral problems such as crowding. Maintaining satisfactory behavioral solutions using conventional programming methods is much less intuitive and difficult to design and maintain. This paper discusses the behavioral problems that the CGF confronts, the generic artificial intelligence model that would solve it, and its applications.