EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

An Intelligent PSFOA–CIFG-LSTM Reinforcement Learning Framework for Dynamic Load Balancing in Edge-Cloud Systems

Neetu Bansla

Journal of Intelligent Decision Making and Information Science · 09.09.2026 · 3 (18s) · 489-522

DOI: 10.17762/jidmis.v3.5743 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Intelligent Decision Making and Information Science, online erschienen am 09.09.2026. Verfasst von Neetu Bansla. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Dynamic load balancing is a critical challenge in cloud-edge computing systems due to continuously fluctuating workloads, heterogeneous resources, and stringent quality-of-service requirements. Traditional scheduling techniques might not be flexible enough to cope with volatile system states or conditions, leading to suboptimal resource use, higher latency, and poor service quality. To solve these problems, the Pheromone-based Starfish Optimization Algorithm (PSFOA), Reinforcement Learning (RL), and Coupled Input-Forget Gate Long Short-Term Memory (CIFG-LSTM) are coupled to form an intelligent hybrid dynamic load balancing framework. Utilizing task profiling and resource monitoring, the architecture defines the workload needs and dynamically monitors the CPU, memory, and network consumption of edge and cloud nodes. PSFOA conducts task-to-node optimization across the globe, and the RL-based scheduler makes task assignments adaptively for the current state of the system. Next, the CIFG-LSTM-based orchestrator learns the temporal workload pattern and carries out tasks at the edge or offloads them to the cloud. A fault tolerance module identifies nodes that are failing and redistributes tasks that require that node's services to nodes that are capable of performing those tasks. The experimental results on the Google Cluster Trace dataset (2019) are presented, which show 98.16% accuracy, 92.72% precision, 99.39% recall, and 95.94% F1 score for the proposed framework. Moreover, it achieves 79.78% load-balancing accuracy, reduces the average latency by 20.14%, gets a CPU load variance of 0.0002, and gets a task migration success rate of 100%. The results show that the proposed PSFOA-RL-CIFG-LSTM approach is effective for load distribution, low latency, adaptive scheduling, and reliable task execution in dynamic cloud-edge computing systems.

Zitieren

APA

Bansla, N. (2026). An Intelligent PSFOA–CIFG-LSTM Reinforcement Learning Framework for Dynamic Load Balancing in Edge-Cloud Systems. Journal of Intelligent Decision Making and Information Science, 3(18s), 489-522. https://doi.org/10.17762/jidmis.v3.5743

Vancouver

Bansla N. An Intelligent PSFOA–CIFG-LSTM Reinforcement Learning Framework for Dynamic Load Balancing in Edge-Cloud Systems. Journal of Intelligent Decision Making and Information Science. 2026;3(18s):489-522. doi:10.17762/jidmis.v3.5743

BibTeX
@article{Bansla2026,
  author = {Bansla, Neetu},
  title = {An Intelligent PSFOA–CIFG-LSTM Reinforcement Learning Framework for Dynamic Load Balancing in Edge-Cloud Systems},
  journal = {Journal of Intelligent Decision Making and Information Science},
  year = {2026},
  volume = {3},
  number = {18s},
  pages = {489--522},
  doi = {10.17762/jidmis.v3.5743}
}
Kostenlos starten →