Tools
Tools
Seenivasagan, Deekshiya (2025) Optimizing Energy Efficiency and Latency in Mobile-Edge-Cloud Systems via PPO, LSTM Caching, and DVFS. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (1MB) | Preview |
Preview |
PDF (Configuration Manual)
Download (1MB) | Preview |
Abstract
The increasing demand for computationally intensive and latency-sensitive applications on mobile devices presents significant challenges in energy efficiency, battery life, and speed. This research project addresses these challenges through smart decisions made by the AI-driven framework for energy-efficient and latency-aware task offloading in mobile-edge-cloud (MEC) environments. The solution integrates three key components: Proximal Policy Optimization (PPO) which is a reinforcement learning model for intelligent task offloading decisions, Long Short-Term Memory (LSTM) network for proactive edge caching, and Dynamic Voltage and Frequency Scaling (DVFS) for adaptive power management. The framework is evaluated and implemented in a simulated environment using CloudSimPlus, with the Python-trained machine learning models has been integrated into the Java based simulator via ONNX Runtime. Various experiments were conducted across eight scenarios in an organized manner by enabling and disabling each optimization component. The results demonstrated that the integrated approach (PPO+LSTM+DVFS) achieves significant improvements in energy consumption, device battery lifetime, and overall task latency when compared to heuristic-based and other baseline approaches. The overall findings demonstrate that combining intelligent offloading, predictive caching, and dynamic power management provides better results for modern MEC systems. This study provides both theoretical understanding and practical guidance for designing intelligent and resource-efficient MEC infrastructures.
Actions (login required)
 |
View Item |