Deep Learning-Based Protocol Stack Optimization in High-Density 5G Environments

Inthiscontext,artificialintelligenceand,morespecifically,machinelearningalgorithmsarebeingdeployedtohelpautomatize,optimize,andinnovate the methods beingapplied. Focusingonthe networkprotocol stack, we consider a DeepLearningapproachtooptimize the coreTCP/IPlayer for 5G environments when the terrestrial network is supporting large numbers of connected devices, running massive machine-type communications and multimedia-enhancedmobilebroadbandapplicationsutilizingnetwork slicing.In particular,consideringaspecificscenariowheredeepreinforcement learning is employed to control and optimize TCP protocol congestion window size, we show that Deep Learning-Based approaches might facilitate TCPoptimizationandavoidcongestionin5Gnetworks,leadingthewaytowardstheimplementationof smart solutions for protocol stack overhead use in dense IoT wireless environments.