TwinFog-MARL: Code, Checkpoints, and Extended Nine-Run Simulation Results for Fog Resource Orchestration
This repository provides the source code, model checkpoints, experimental datasets, statistical analyses, and visualization scripts supporting the TwinFog-MARL study on energy–latency–reliability trade-offs in fog resource orchestration. The extended validation campaign comprises nine independent training initializations, each trained for 100 episodes of 60 simulation steps, with shared held-out workload evaluations. The repository includes comparisons against an adapted FPLB implementation, least-loaded scheduling, and round-robin scheduling; an evaluation-time cloud-redirection ablation; and sensitivity analysis of simulated cloud-energy costs. The archive also contains run-level statistical comparisons, exact paired sign-flip permutation tests, and reproducibility instructions. Energy measurements are simulator-based, cloud-cost sensitivity scenarios are hypothetical rather than hardware-calibrated, and the no-redirect policy was not separately retrained. Keywords: Fog Computing; Multi-Agent Reinforcement Learning; PPO; Task Scheduling; Energy Efficiency; SLA; Reproducibility
Authors
- Waleed Mahmoud Ead (ORCID: https://orcid.org/0000-0003-2969-2269)
Institutions
- Beni-Suef University (EG)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-18
- DOI
- https://doi.org/10.5281/zenodo.24353337
- Primary Topic
- IoT and Edge/Fog Computing
- Type
- article
- Field-Weighted Citation Impact
- 0.00