Análise de mobilidade e um Autoencoder Robusto

Abstract

Statistical channel modeling plays an important role in the development of commu nication networks. With the advent of 5th generation of mobile networks (5G) and 6th generation of mobile networks (6G), it is necessary to use generalist models, since networks are expected to be increasingly diversified in terms of connected devices and with greater need for resources and efficiency. A promising paradigm for modern networks is artificial intelligence (AI), with the role of optimization, integration and management at various levels. This work seeks to evaluate a general α-µ fading model affected by Gamma sha dowing in a random waypoint model (RWP) mobility scenario for different propa gation environments and physical network topologies. New expressions were ob tained for probability density function (PDF), cumulative distribution function (CDF), average symbol error probability (ASEP), outage probability (OP) and capacity. Then, the application of a communication system based on dense neural network (DNN) as an autoencoder (AE) in the proposed channel is investigated. With only the knowledge of the channel samples, the AE obtained a performance similar to traditional modulations and proved to be robust for channel variations.