Computer modeling of intelligent agent behavior is a concern to many researchers in the fields of cognitive science, artificial intelligence, and psychology. The Soar community is particularly interested in developing a model which encompasses a unified theory of cognition [Soar]. To this end, Soar researchers are interested in modelling agents that operate in challenging environments. Dynamic environments which require the application of a fair amount of domain knowledge offer a diverse set of problems that must be addressed in developing agents which simulate intelligent behavior. These problems require the development of a number of cognitive facilities in order to successfully simulate agent behavior and the unified approach of Soar is helpful in merging these facilities into a coherent whole.
One environment which provides these challenges is the domain of air-to-air combat. In this domain, fighter pilots must make quick decisions concerning enemy aircraft in the service of completing a mission. Developing an agent to operate in such an environment is much too costly and the natural testbed for such development is a simulator. The simulator chosen to aid the development of a Soar pilot is the ModSAF system [ModSAF].
ModSAF provides a platform for research into the control of all kinds of computer generated forces. In essence, ModSAF simulates the operation of Distributed Interactive Simulation (DIS) compatible vehicles. These vehicles can be directed by software-controlled agents or human beings. By using ModSAF, researchers can focus their work on the development of believable agents rather than on vehicle simulation issues, such as motion dynamics and DIS networking. Our work deals with the problems of interfacing artificially intelligent agents modeled using the Soar system to ModSAF. The module which supports the connection between the two systems is called the Soar/ModSAF Interface (SMI).