RedTops: real-time energy-aware dynamic task offloading via federated mountain gazelle optimisation in SDN-enhanced edge computing

作     者：Al Aghbari, Zaher Khedr, Ahmed M. Ahmed, Naveed Girija, Shini Baker, Thar 

作者机构：Department of Computer Science University of Sharjah Sharjah United Arab Emirates School of Architecture Technology and Engineering University of Brighton BrightonBN2 4GJ United Kingdom 

出 版 物：《Neural Computing and Applications》 (Neural Comput. Appl.)

年 卷 期：2025年第37卷第19期

页      面：13795-13833页

核心收录：

学科分类：08[工学] 0835[工学-软件工程] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

主　　题：Federated learning 

摘      要：Offloading computational tasks is vital for real-time applications on mobile devices with limited resources. Mobile edge computing (MEC) is deemed a solution that puts computational resources closer to users. Nevertheless, there are many associated concerns during the offloading procedure (i.e., privacy, delay, and high energy consumption). Federated learning (FL) has been considered a solution to address MEC’s data privacy issues;however, it comes with its own resource consumption issues. To address these issues, this paper proposes a distributed learning paradigm inspired by FL. We propose an optimisation technique for offloading computational tasks that aims to reduce both total delay and energy consumption by using the mountain gazelle optimisation algorithm, which shows it can reduce both delay and energy consumption in dynamic situations. Additionally, an improved variant known as the improved mountain gazelle optimiser is integrated into a distributed SDN controller architecture to create an offloading policy model for optimal edge node selection. We also present a new SDN-enabled edge computing architecture that achieves the best task distribution through task offloading using federated mountain gazelle optimisation (RedTops). Energy usage, delay, and bandwidth are considered by RedTops, which successfully addresses high training costs, dependability issues, and privacy concerns in MEC. Based on the outcomes of five extensive simulations, RedTops is more energy-efficient and faster at completing tasks than four state-of-the-art offloading methods (DDLO, DROO, DRL without TL and SDN, and DTRL). © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2025.