Adaptive Reinforcement Learning-Based Framework for Energy-Efficient Task Offloading in a Fog-Cloud Environment.

Area of Science:

• Computer Science
• Artificial Intelligence
• Distributed Computing

Background:

• The proliferation of Internet of Things (IoT) devices increases computational demands, leading to significant energy consumption concerns in fog-cloud environments.
• Energy-efficient task offloading is critical for time-sensitive tasks on resource-constrained IoT devices.

Purpose of the Study:

• To propose a novel reinforcement learning-based framework, Adaptive Q-learning-based Energy-aware Task Offloading (AQETO), for dynamic energy management in fog-cloud networks.
• To address the challenge of energy-efficient task offloading for time-sensitive IoT tasks while considering delay tolerance and deadline requirements.

Main Methods:

• AQETO dynamically determines fog node energy states using Q-learning based on workload fluctuations.
• The framework prioritizes the allocation of urgent tasks to minimize processing delays.
• Computational resources are allocated considering IoT task delay tolerance and deadline satisfaction.

Main Results:

• AQETO effectively minimizes energy consumption in fog nodes.
• The proposed framework significantly reduces task delays.
• System efficiency is maximized through optimized resource allocation and task prioritization.

Conclusions:

• AQETO provides an effective solution for energy-efficient task offloading in fog-cloud environments.
• The framework successfully balances energy consumption, task delay, and system efficiency.
• Reinforcement learning, specifically Q-learning, is a viable approach for dynamic resource management in IoT networks.