Human-Directed Learning for Unmanned Air Vehicle Systems in Expeditionary Operations

Effectively targeting and engaging time sensitive targets (TST) is an important part of modern warfare. However, using UAVs for this task is highly challenging because of difficulties in optimally retargeting the UAVs while taking into consideration multiple variables, such as time to target, remaining fuel, uncertainties in the location and type of the TSTs, avoidance of threat areas etc. Systems for UAV control that can learn to engage TSTs with knowledge-based reasoning akin to human judgment and expeience would be highly valuable for decreasing the cognitive load on the human operator, decreasing the time for decision making for increasing the reliability and robustness of the decision making process. We propose such a system called UAV-Guide that observes and learns from the actions an expert human operator takes while targeting and engaging TSTs. UAV-Guide combines the technologies of knowledge representation, learning and human factors to address the problems of representing and reasoning with uncertain knowledge, learning from observations and reducing cognitive overload. Careful consideration has been given in UAV-Guide to issues critical to user adoption, such as providing facilities for the user to direct the learning process and to override what is automatically learned, when necessary. Phase I prototype development will provide a solid foundation for complete implementation in Phase II.