Task Offloading and Resource Allocation Strategy Based on Deep Learning for Mobile Edge Computing.

Area of Science:

• Computer Science
• Artificial Intelligence
• Distributed Computing

Background:

• Mobile Edge Computing (MEC) faces challenges with inefficient computation offloading and uneven resource distribution.
• Optimizing task completion time and energy consumption is crucial for MEC system performance.

Purpose of the Study:

• To propose a novel deep learning-based strategy for task offloading and resource allocation in MEC.
• To address the limitations of unreasonable computation offloading and uneven resource allocation in multiuser multiserver MEC environments.

Main Methods:

• A new objective function was designed, integrating calculation and communication models to minimize task completion time and device energy consumption under delay constraints.
• A multiagent reinforcement learning system was employed, with system benefits and resource consumption defined as rewards and losses.
• The Dueling-DQN algorithm was utilized to determine the optimal resource allocation strategy.

Main Results:

• The proposed strategy achieved the best performance with a learning rate of 0.001 and a discount factor of 0.90.
• Demonstrated a 52.18% reduction in energy consumption and a 34.72% decrease in task completion time.
• Outperformed other comparison strategies in terms of computational load and energy savings.

Conclusions:

• The deep learning-based task offloading and resource allocation strategy effectively optimizes MEC system performance.
• The Dueling-DQN algorithm provides a robust solution for complex resource management in MEC.
• Significant improvements in energy efficiency and task completion time validate the proposed approach.