Modelling, learning and prediction of complex radar emitter behaviour

This paper adapts and extends the previously published hierarchical modelling approach of multifunction radars as systems that speak a language. We propose Long Short-Term Memory neural networks for learning the emitters' grammar and predicting their emissions. The approach is demonstrated using simulations of an airborne multifunction radar with three different resource management techniques of varying complexity. A comparison with simple prediction strategies shows that a huge improvement in accuracy can be achieved by using Long Short-Term Memory networks for predicting complex radar emitter behaviour.