A Survey on Adaptive and Decentralized Task Scheduling in Fog Computing Environments

The rapid growth of Internet of Things (IoT) applications has created strong requirements for low-latency response and reduced energy consumption during data processing. In practical deployments, relying solely on distant cloud infrastructure often fails to meet these constraints. Fog computing partially addresses this issue by relocating computation closer to end devices. However, task scheduling in decentralized fog environments is far from trivial. The coexistence of heterogeneous computing nodes, fluctuating workloads, and occasional node mobility limits the effectiveness of static or rule-based scheduling strategies.

In this study, we introduce an adaptive scheduling framework designed specifically for decentralized fog architectures that must operate under continuously changing system conditions. Reinforcement learning is adopted to enable fog nodes to gradually learn effective task-allocation decisions by observing current network states and resource availability. At the same time, federated learning is incorporated to facilitate coordination among distributed nodes while avoiding direct data sharing, thereby addressing privacy concerns. The proposed framework aims to minimize task execution latency, improve overall energy efficiency, and support scalability without relying on centralized orchestration. Its performance will be evaluated through experiments conducted under realistic fog computing scenarios.