RAPID WAVEFORM DESIGN THROUGH MACHINE LEARNING

In this paper, we discuss the feasibility of the novel application of recurrent neural networks (RNN) in designing lowlatency, near-optimal radar waveforms in dynamical environments. Traditional approaches to adaptive radar waveform design typically require cumbersome optimization routines and highly specialized solvers that can be slow to converge. In an effort to decrease the time of convergence, while still being robust to dynamic environments and practical implementation concerns, we provide results with use of RNN tools. In these initial trials, we achieve waveform design results with comparable characteristics of the Error Reduction Algorithm.