Radar Data in a Deep Learning Representation (RDDLR)

Modern radars operate in a variety of modes and employ different waveforms to enable them to achieve improved performance in a variety of conditions. Traditional approaches to detection and exploitation of these emitters often produce erroneous results due, in part, to solid state transmit electronics and the sheer variety of modes and waveforms a single radar may emit. This SBIR is focused on improving the representation of the threat data to correctly detect and characterize the threat emitters. Traditionally platforms are provided a static table-based library for mission and reprogramming data, tailored to the Area of Operations. Traditional approaches, however, fail to address the dynamics in the threat environment encountered during a mission, the distribution of newly formulated signal characterization between platforms, and the agility of modern radars to transition states. To meet this need, EXP proposes Radar Data in a Deep Learning Representation (RDDLR), a transition to a new compact emitter waveform data representation along with a shift in platform library organizational principles, management, and content. RDDLR supplements the database with new learned features, while updating the data with new emitters or new modes of existing emitters in real time.