Predicting wireless coverage maps using radial basis networks

Accurate assessment of the wireless coverage of a station is a critical step toward deploying more base stations in Ultra Dense Networks, and it is considered as one of the key features of the 5G networks. Quickly and efficiently determining the reception coverage of transmitters becomes a complicated problem when interfering transmitters are introduced to the scenario. It becomes increasingly more complicated when the transmission powers of those transmitters are not uniform. Artificial Neural Networks are the most suitable learning algorithms for recognizing and predicting non-linear patterns. In particular, a Radial Basis Network is a type of Artificial Neural Network which typically uses a Gaussian kernel as an activator as opposed to a sigmoid function. In this paper, we suggest using Radial Basis networks in order to predict coverage maps. We show how it is possible to train the Radial Basis Network to generate coverage maps based on samples and we check the accuracy level of the learning process on a test set. Using Radial Basis Network can improve the cellular coverage prediction and therefore it can enable a more efficient spectrum allocation.