Advancing the State of the Art in Artificial Intelligence for Simulation Training

The USMC has an immediate need for a framework and infrastructure with which state-of-the-art AI technologies can be integrated to replace simplistic CGF behaviors in virtual simulation platforms such as VBS2. SoarTech, along with our partner Aptima, proposes to apply our vast expertise in the design, development and integration of artificial intelligence technologies, bringing it to bear to help develop more realistic entity-level scenarios for USMC simulated training. Most crucial to our work will be the design and development of a full infrastructure, called SERUM (Simulated training Exercises with Robust Unmanned Models) to support advanced AI for the USMC"s simulation technology portfolio. This infrastructure will allow for robust, unmanned entities (Soar-based and otherwise) to exist in both constructive (e.g. OneSAF, JSAF) and virtual (e.g. VBS2) simulation environments simultaneously, and will also include the necessary hooks and mechanisms to allow entities to both perceive the world and make actions within it. Using SERUM, we will then develop prototype robust intelligent agents (and apply existing agents) that can serve as robust, tactically viable role-players for a particular domain (e.g. JTAC training) that require minimal babysitting and maintenance on the part of scenario operators.