Sparse array reconfigurability for source identification and angle estimation in cognitive sensing

This paper considers the role of sparse arrays in multi-function Radio Frequency (RF) sensor node where both the source angular direction and the source signal characteristics are of interest. The sparse array configuration is determined by the mode of operation within the cognitive Perception-Action Cycle (PAC) of the underlying sensing problem. Where the perception mode in PAC uses a fully augmented sparse array to estimate the directions and powers of all sources in the field of view (FOV), the Action mode determines the optimum common sparse configuration that yields the highest average signal-to-interference and noise ratio (SINR) for isolations and characterizations of all source signals. Since both modes require fast computations and reactions, we apply Deep Neural Networks (DNNs) to learn the common array configuration and employ fast iterative FFT-based techniques for direction of arrival (DOA) estimation. The performance of the PAC is evaluated in terms of accurate signal recovery measured by bit error rates.