15.097 Student Project: Operator-Defined SUAV Classification Tree Algorithms

The growing divide between the potential of fully automated teams of Unmanned Aerial Vehicles (UAVs) and the very limited autonomy seen in operational use deserves attention. This paper proposes a limited step to realize increased autonomy use in operational settings by applying decision tree learning algorithms to UAV operator’s flight patterns. These algorithms learn and then propose UAV trajectories based on the individual preferences and patterns of each operator. This approach is distinct from previous methods in that it seeks to tailor autonomy suitable to each operator rather than seek a panacea for all. We show that C4.5 classification trees have numerous advantages over competing algorithms, and limited experimental results have provided great promise for future research.