Paper Key : IRJ************430
Author: Dr. A. Kanthimathinathan,Dr.s.saravanan,Dr. P. Anbalagan
Date Published: 14 Nov 2024
Abstract
The cloud-edge-terminal framework would improve the computational capability, response time, and resource exploitation of computations about intricate operations. However, efficient scheduling is a major research issue because these environments are highly diverse and dynamic. This paper articulates new task scheduling algorithms for adaptive Cloud Federated Systems based on Federated Deep Reinforcement Learning (DRL) in response to these challenges. As the learning process is forced to be distributed and the model can be trained across multiple network nodes, federated DRL retains data privacy and helps adapt to network conditions. The introduced algorithms on the cloud side shall enable task distribution according to current resource availability and workloads of cloud edge and terminal devices. The analysis of experimental data suggests that these adaptive scheduling algorithms greatly impact the overall improvements in task completion time, lower latency rate, and increase system efficiency compared with conventional centralized scheduling methods. It is for this reason that this study highlights the feasibility of federated DRL as a novel approach to intelligent task scheduling in cloud-edge-terminal collaborative networks.