Reinforcement Learning based Waveform Design for Cognitive Imaging Radar

The introduction of artificial intelligence technology into cognitive radar is becoming a hot research direction. Cognitive radar can feed back the prior information to the transmitter to form a cognitive loop and then improve the detection performance. Designing transmitting waveform plays a key role in cognitive loop. In this paper, one cognitive imaging waveform selecting reinforcement learning (RL) approach is proposed by combining the waveform design and deep reinforcement learning (DRL). Then it focuses on the problem in inverse synthetic aperture radar (ISAR) sparse imaging that evaluating the number of randomly transmitting subcarriers and then adaptively adjusting the transmitting waveform. Simulation results showed that the proposed approach preliminarily realized the purpose of evaluating and adjusting.