Detección de Nodos en Zonas Ocultas en redes LAA a través de Aprendizaje Automático Supervisado

LTE operation in unlicensed spectrum bands, based on Licensed-Assisted Access (LAA), is considered as an option to increase the capacity of 4G wireless networks. These solutions use a Listen Before Talk (LBT) protocol that enables the eNodeB (eNB) to access the medium opportunistically, avoiding collisions from/to other eNBs. However, the hidden node problem must be addressed in LAA networks to reduce or prevent the degradation of the network.  The efficiency of the LTE-LAA system will improve by identifying hidden nodes and after deciding if user equipment (UE) affected by the hidden condition should remain or should change from unlicensed to licensed band. In this work, we use two supervised machine learning algorithms to determine if UEs located in the border cell are affected by hidden nodes. We trained a logistic regression and a neural network with standard parameters obtained from UE to detect when UE is affected by collisions due to the presence of hidden nodes. The results show that the neural network has a perfect performance as a detector of UE facing hidden nodes  Keywords:  LAA, LBT, Hidden Node, RSRI, RSRP, RSRQ.