5G Converged Network Resource Allocation Strategy Based on Reinforcement Learning in Edge Cloud Computing Environment

Area of Science:

• Computer Science
• Artificial Intelligence
• Telecommunications Engineering

Background:

• Mobile Edge Computing (MEC) servers face challenges processing intensive, long-period task data due to limited computing power and resources.
• The increasing demand for mobile services necessitates efficient resource management in edge cloud environments.

Purpose of the Study:

• To propose a 5G converged network resource allocation strategy for edge cloud computing environments.
• To address the limitations of local computing power in MEC by offloading tasks.
• To minimize average system task response time and total energy consumption.

Main Methods:

• Developed a multi-MEC server and multi-user mobile edge system.
• Modeled task offloading and resource allocation as a Markov decision process.
• Utilized a deep Q-network algorithm for optimal resource allocation scheme determination.
• Experimentally analyzed the proposed strategy using the TensorFlow learning framework.

Main Results:

• The proposed strategy effectively reduces task delay.
• Experimental results demonstrated improved resource utilization.
• With 110 users, the final energy consumption was approximately 2500 J, showcasing efficiency.

Conclusions:

• The reinforcement learning-based 5G converged network resource allocation strategy is effective for MEC environments.
• The strategy successfully balances task offloading and resource allocation to optimize performance and energy consumption.
• This approach offers a viable solution for enhancing MEC capabilities in handling intensive mobile computing tasks.