Intelligent packet error prediction for enhanced radio network performance
Loading...
URL
Journal Title
Journal ISSN
Volume Title
Sähkötekniikan korkeakoulu |
Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.
Authors
Date
2019-08-19
Department
Major/Subject
Control, Robotics and Autonomous Systems
Mcode
ELEC3025
Degree programme
AEE - Master’s Programme in Automation and Electrical Engineering (TS2013)
Language
en
Pages
83+1
Series
Abstract
In cellular communication systems, for example 4G and 5G, quite often data packets (in user-plane payload) fail to successfully deliver to the user equipment (UE). Because upon failure, a re-transmission of the data packet is required by the network, these failed data packets introduce latency to the network. In some applications, such latency might be tolerable by the UE, but in applications that require ultra reliable low latency communication (URLCC), time latency becomes a critical issue. In order to cope with this issue, typically wireless networks rely on re-transmissions upon receiver request or use naïve approach like packet duplication to transmit data packets more than once to ensure successful transmission of at least one data packet without any error. In this thesis, we explore the feasibility of designing an intelligent solution to this issue by using network data with machine learning and neural networks to predict if a data packet would fail to transmit in the next transmission time interval (TTI). Our research includes a detailed systematic study on which radio parameters to choose from the raw data (log files) and data preprocessing. From our experiments we also determine how many past values of these radio parameters can be useful to predict the packet failure in the next TTI. Moreover, we enlist the network parameters useful to make such a prediction and compare their contribution in the model. Finally, we show that an intelligent packet error prediction can be done using machine learning that forecasts the packet failure in the next TTI with sufficient accuracy. We compare the performance of different machine learning algorithms and show that boosted decision trees (XGBoost) perform the best on the given dataset. Compared to naïve approaches used in cellular communication to avoid packet failures, our solution based on intelligent packet error prediction indicates promising practical applications in cellular network for enhanced radio network performance, particularly in URLLC.Description
Supervisor
Särkkä, SimoThesis advisor
Korpi, DaniKeywords
packet, error, prediction, with, machine, learning