Deep learning-based user location estimation in SDN networks

Abstract

Indoor localization systems, which employ a variety of sensors on mobile phones, have advanced due to their popularity. Handover is needed to maintain communication when a mobile user travels between base stations (BS), particularly between cells. We presented and tested the CNN model for predicting SDN network users' upcoming positions using a portion of their actual mobility traces. We recommend incorporating convolutional neural networks to improve input data representation. The number of elements of each array antenna, the pencil parameter, the number of frequency reference points, and the distances between them based on the maximum delay were determined by dividing the urban telecommunication SDN network into 16 sub-sections and using 3 base stations. We also thought the mobile channel's core frequency was crucial to determining the user's whereabouts. Thus, we obtained 1,000 test samples for the requested data set. We developed deep learning architecture to estimate user locations in the next step. Using the Xception network architecture, Adam solver, and trainbr training function, we estimated user locations. Model implementation results show that the CNN model with 3700 repetitions assessed the user's location with 98.65% accuracy and 0.15 RMSE. R=0.91934 shows the results are associated.