📝 SummarXiv

Abstract

In this work, we develop centralized and decentralized signal fusion techniques for constant false alarm rate (CFAR) multi-target detection with a cognitive radar network in unknown noise and clutter distributions. Further, we first develop a detection statistic for co-located monostatic MIMO radar in unknown noise and clutter distributions which is asymptotically CFAR as the number of received pulses over all antennas grows large, and we provide conditions under which this detection statistic is valid. We leverage reinforcement learning (RL) for improved multi-target detection performance, where the radar learns likely target locations in a search area. These results are then generalized to the setting of cognitive radar networks, where radars collaborate to learn where targets are likely to appear in a search area. We show a fundamental tradeoff between the spatial and temporal domain for CFAR detection in unknown noise and clutter distributions; in other words, we show a tradeoff between the number of radar antennas and the number of temporal samples. We show the benefits and tradeoffs with centralized and decentralized detection with a network of cognitive radars.