Task offloading optimization in mobile edge computing based on a deep reinforcement learning algorithm using density...

Area of Science:

• Computer Science
• Artificial Intelligence
• Network Engineering

Background:

• Mobile edge computing (MEC) enhances performance by offloading tasks from mobile devices to edge servers, enabling low-latency services.
• Inefficient task offloading in dynamic edge environments can lead to resource contention and suboptimal allocation due to limited bandwidth.
• Existing methods struggle with efficient resource management and task prioritization in complex MEC networks.

Purpose of the Study:

• To propose a novel deep reinforcement learning method for intelligent task offloading decision-making in MEC.
• To enhance computing performance and resource allocation efficiency in mobile edge environments.
• To address challenges of limited bandwidth and dynamic network conditions in MEC.

Main Methods:

• The proposed Density Clustering and Ensemble Learning training-based Deep Reinforcement Learning (DCEDRL) method utilizes multiple deep neural networks.
• Ensemble learning combines predictions from multiple models for robust decision-making.
• An optimized density clustering method classifies tasks by characteristics, improving scheduling and resource allocation.

Main Results:

• The DCEDRL method demonstrated a reduction in task backlogs exceeding 21% compared to baseline algorithms.
• DCEDRL improves the adaptability and robustness of the system through real-time sampling strategy adjustments.
• The method effectively manages computing resources and prioritizes tasks in mobile edge environments.

Conclusions:

• DCEDRL offers a significant improvement in task offloading efficiency for mobile edge computing.
• The integration of density clustering and ensemble learning enhances resource allocation and system performance.
• This approach provides a robust solution for managing tasks in dynamic and resource-constrained MEC networks.