Utility-Driven End-to-End Network Slicing for Diverse IoT Users in MEC: A Multi-Agent Deep Reinforcement Learning...

Area of Science:

• Computer Science
• Telecommunications Engineering
• Artificial Intelligence

Background:

• Mobile Edge Computing (MEC) enhances Quality of Service (QoS) by reducing latency.
• Deploying cloudlets for MEC faces challenges in network slicing due to uneven traffic and diverse resource needs.
• Efficient resource allocation and prediction of request types are critical for MEC performance.

Purpose of the Study:

• To develop an intelligent network slicing approach for Mobile Edge Computing.
• To optimize resource allocation and utility in MEC environments with fluctuating demands.
• To address challenges in managing distributed cloudlets and network slicing.

Main Methods:

• Utilized T distribution for predicting common and urgent user requests.
• Formulated the problem as a multi-agent Markov decision process (MDP).
• Introduced a multi-agent soft actor-critic (MAgSAC) algorithm for intelligent Virtual Network Function (VNF) instance management.

Main Results:

• The MAgSAC algorithm optimizes overall utility by 30%.
• Achieved a 12.4% reduction in energy consumption costs.
• Reduced execution time by 21.7% compared to the MAA3C benchmark.

Conclusions:

• The proposed MAgSAC approach effectively balances resource allocation, revenue, energy costs, and latency.
• Intelligent VNF instance activation/deactivation prevents resource wastage in MEC.
• MAgSAC offers a superior solution for network slicing optimization in edge computing environments.