Deep Learning Sparse Array Design Considering Binary Switching and Missing Coarray Lags

Sparse array reconfigurability via fast switching is pivotal to a realizable perception-action cycle of cognition in dynamic radio frequency (RF) environments. In this paper, we propose a deep learning (DL) sparse array design method using binary switching per RF chain for optimum beamforming that maximizes the signal-to-interference-and-noise ratio (MaxSINR). We consider two binary switching strategies and design the arrays to incorporate missing spatial data correlation values. Through design examples, we demonstrate that the DL method can effectively learn the optimum sparse arrays for MaxSINR in the presence of a missing correlation lag.