Dynamic Spectrum Assignment for Land Mobile Radio with Deep Recurrent Neural Networks
Humphrey Rutagemwa, Amir Ghasemi, Shuo Liu · 2018
In this paper, measurements from a spectrum awareness system are used to study the application of machine learning methods in dynamic spectrum assignment for Land Mobile Radio (LMR). Specifically, a deep recurrent neural network is used to learn the time-varying distributions of users' traffic, which in turn, help to determine the best spectrum assignment and sharing strategies in LMR bands. Using RF data, network simulations are conducted to validate and evaluate the suitability of the chosen methodology. It is shown that the deep learning approaches have great potential for characterizing spectrum usage patterns and facilitating spectrum assignment decisions in dynamic wireless environment.