Mixed Initiative Machine for Instructed Computing: MIMIC II

In less than a decade, UAVs have transitioned from a curiosity on the battlefield to a core capability for intelligence and attack. However, the intelligence that drives tactical decision making and flight control remains firmly lodged in the heads of expert operators. As more UAVs and other unmanned vehicles enter the battlespace, intelligent, autonomous UAVs will become a necessity. To address this challenge, Aptima, Boeing, and the Cognitive Engineering Research Institute propose to (1) conduct human subjects experiments that illuminate the potential for humans to convey their knowledge of autonomous tactical decision making and control to UAVs, and to (2) prototype training support technology and machine learning algorithms that enable UAVs to learn from expert UAV operators. We call the proposed capability MIMIC: Mixed Initiative Machine for Instructed Computing. MIMIC is a hybrid model that integrates psychological learning theory and rapid machine learning algorithms to enable human operator to teach the UAV new tactical and control behaviors through a unique user interface to a UAV mission simulator. MIMIC will enable UAVs to infer the mission threats and opportunities and to dynamically optimize tactical decisions and control actions when the user cannot do so or obligates the UAV to do so.