Spectral Prediction and Notching of RF Emitters for Cognitive Radar Coexistence

The increasing demand for radio frequency (RF) spectrum requires radars to coexist with interfering RF emitters. Innovations in waveform design paired with dynamic spectrum access techniques allow for efficient spectrum sharing for radar. This work evaluates a recently developed real-time implementation of spectrum sharing cognitive radar using commercial-of-the-shelf (COTS) hardware. Prediction of coexisting RF emitter frequencies informs the design of spectrally notched FM noise waveforms on transmit. Waveform notches are optimized on a pulse-to-pulse basis while accounting for the zero-order hold model inherent to lower fidelity digital-to-analog converters to ensure desired reconstruction. The radar system employs cognition to learn and predict RF emitter activity via a stochastic model-based approach. Initially, passive spectrum observations are used to estimate a stochastic model which is then exploited to predict the likelihood of future RF activity. The benefits and limitations of this sense-predict-and-notch (SPAN) approach are evaluated using a set of synthetic interference scenarios in real-time.