Resource Adaptive Automated Task Scheduling Using Deep Deterministic Policy Gradient in Fog Computing

Abstract

The rapidly increasing complexity of Internet of Things applications and the exponential growth in data generation pose significant challenges in terms of latency and network capacity constraints, especially in cloud computing. Fog computing carries have emerged as an effective solution by decentralizing data processing, reducing latency, and bringing computation closer to the data sources. This paper presents a novel adaptive scheduling framework based on the DDPG algorithm for task scheduling optimization in fog computing environments. Our framework is based on DDPG, a reinforcement learning algorithm well suited for continuous action spaces, adapting scheduling strategies to real-time changes in task demands and resource availability. Such capability allows for highly accurate decisions to be made in real time, which is of particular importance for latency-sensitive applications such as autonomous vehicles, remote healthcare, and industrial automation. The contributions of this paper include the development of an adaptive scheduling framework that can support sequential, parallel, and dependency-based scheduling. This framework improves several critical performance metrics by 30%, reduces makespan, reduces fault tolerance by 25%, and improves system scalability and reliability by 20%. Data processing localization through our approach reduces bandwidth and latency usage by up to 40% and improves data privacy and efficiency in data management. Simpy simulated this work using Google Cloud Jobs datasets, and the results support the promising combination of advanced machine learning and fog computing. This research demonstrates the transformative impact of deep reinforcement learning in improving task scheduling in distributed computing environments, laying a strong foundation for further research aimed at harnessing the full capabilities of fog computing for the ongoing demands of IoT applications.