Federated Deep Reinforcement Learning-Based Task Offloading and Resource Allocation for Smart Cities in a Mobile Edge...

Area of Science:

• Computer Science
• Artificial Intelligence
• Network Engineering

Background:

• Mobile Edge Computing (MEC) is crucial for Industry 4.0 and smart cities, enabling computation offloading.
• Offloading intensive tasks to MEC or cloud servers raises privacy concerns due to raw data migration.
• Federated learning offers a privacy-preserving approach to enhance training performance.

Purpose of the Study:

• To address privacy disclosure issues in MEC task offloading.
• To minimize energy consumption for Internet of Things (IoT) devices.
• To optimize joint task offloading and resource allocation under delay constraints.

Main Methods:

• Formulation of a joint optimization problem for task offloading and resource allocation.
• Development of a two-timescale federated deep reinforcement learning algorithm (FL-DDPG).
• Utilizing the Deep Deterministic Policy Gradient (DDPG) framework.

Main Results:

• The proposed FL-DDPG algorithm significantly reduces energy consumption for IoT devices.
• The algorithm effectively manages task offloading and resource allocation.
• Simulation results validate the algorithm's performance in minimizing energy usage.

Conclusions:

• Federated learning integrated with MEC effectively mitigates privacy risks.
• The FL-DDPG algorithm provides an efficient solution for energy-efficient task offloading in smart cities.
• This approach supports the development of secure and sustainable intelligent manufacturing and smart city ecosystems.