Joint optimization of sonar waveform selection and sonobuoy placement

We propose a method to jointly optimize placement of a sonobuoy field and selection of appropriate active sonar waveforms in a complex undersea environment, using deep reinforcement learning. To this end, we develop a custom simulation environment and train an online algorithm that uses updated information as sonobuoys are placed and sonar measurements made. The approach may be termed multiple bistatic, though it shares features with multistatic approaches. We develop benchmarks using conventional fixed sonobuoy placement patterns together with fixed composite sonar pulse trains and present results that show our new approach can outperform the best conventional combination of fixed placement pattern and pulse train.