2019 I/ITSEC

Sessions Speakers Exhibitors Products Press Releases Show Specials Floor Plan Buzz

Man-Machine Interoperation in Training for Large Force Exercise Air Missions (Room 320A)

The United States Air Force strives to maximize human abilities in highly complex operational environments, and artificial intelligence (AI) affords opportunities to transform voluminous data into meaningful information to support human decision making. A Mission Analysis and Review System (MARS) was developed to explore how AI can automate current mission debrief processes and to visualize that information in a mission-specific context. The current effort explored the development of AI to assist Air Force commanders in evaluating the mission performance of a Large Force Exercise (LFE), which affords pilots the chance to hone their abilities to execute their individual role within a mission that may include dozens of aircraft.   In an earlier but related effort, the research team developed machine intelligence to automatically label mission phases of a two-ship strike formation using entity state data of aircraft flown by human pilots in simulation. In the current effort, an LFE with 18 friendly aircraft was simulated using the Joint Semi-Automated Forces (JSAF) simulation engine. Models were created to score both individual aircraft behavior as well as overall mission objective success. Templates were used to determine if acceptable levels of key mission objectives are being estimated and evaluated. By enumerating the propositions included in the three temporal behaviors in the classification model, the behaviors the model deems necessary for evaluating the execution as acceptable were interpreted. Results showed that mission phases and their objectives could be correctly classified with an accuracy of .92 to .96 using a technique where mission objectives were encoded in a linear temporal logic (LTL) format.   The findings suggest that AI can be used to make meaning of raw data for use by commanders to support LFE planning and debrief. The effort described is the first known application of machine intelligence to automatically score mission performance for