Machine-Learning-Based Parallel Genetic Algorithms for Multi-Objective Optimization in Ultra-Reliable Low-Latency WSNs

Date

2018-12-10

ORCID

Journal Title

Journal ISSN

Volume Title

Publisher

Institute of Electrical Electronics Engineers Inc

item.page.doi

Abstract

Different from conventional wireless sensor networks (WSNs), ultra-reliable and low-latency WSNs (uRLLWSNs), being an important application of 5G networks, must meet more stringent performance requirements. In this paper, we propose a novel algorithm to improve uRLLWSNs' performance by applying machine learning techniques and genetic algorithms. Using the K -means clustering algorithm to construct a 2-tier network topology, the proposed algorithm designs the fetal dataset, denoted by the population, and develops a clustering method of energy conversion to prevent overloaded cluster heads. A multi-objective optimization model is formulated to simultaneously satisfy multiple optimization objectives including the longest network lifetime and the highest network connectivity and reliability. Under this model, the principal component analysis algorithm is adopted to eliminate the various optimization objectives' dependencies and rank their importance levels. Considering the NP-hardness of wireless network scheduling, the genetic algorithm is used to identify the optimal chromosome for designing a near-optimal clustering network topology. Moreover, we prove the convergence of the proposed algorithm both locally and globally. Simulation results are presented to demonstrate the viability of the proposed algorithm compared to stateof-the-art algorithms at an acceptable computational complexity.

Description

Full text access is available from the publisher. Use the Link to Article

Keywords

Clustering algorithms, Telecommunication, Machine learning, Mathematical optimization, Electric network topology, Wireless sensor networks

item.page.sponsorship

NPRP under Grant 8-627-2-260 from the Qatar National Research Fund

Rights

Open Access Publishing Agreement, ©2018 IEEE

Citation