کنفرانس بین المللی فناوری اطلاعات و دانش

صفحه اصلی / شانزدهمین کنفرانس بین المللی فناوری اطلاعات و دانش

Distributed Deep Reinforcement Learning for Energy-Efficient and Low-Latency Load Balancing in Mobile Edge Computing

نویسندگان :

Pooria Azizi¹ Siavash Khorsandi²

1- Amirkabir University of Technology 2- Amirkabir University of Technology

کلمات کلیدی :

Mobile Edge Computing،Load Balancing،Double Deep Q-Network،Prioritized Experience Replay،Reinforcement Learning،Beyond 5G Networks،Energy Efficiency

چکیده :

The rapid growth of the Internet of Things (IoT) and next generation networks has increased real time data demand, requiring improved Quality of Service (QoS) and energy efficiency. Edge computing, through fog nodes near users, reduces latency and enhances responsiveness. However, dynamic workloads cause resource imbalance, necessitating intelligent load balancing. This study proposes a multi agent reinforcement learning (MARL) framework for optimizing fog node performance. Two models are evaluated: one with state sharing among neighbors and another independent model. Simulations show a 19\% improvement under static conditions for non sharing and an additional 3.2\% gain under mobility above 50\% for the sharing model. The approach effectively adapts to device mobility while maximizing resource utilization and minimizing energy consumption in next generation edge networks.