کنفرانس مهندسی برق ایران

صفحه اصلی / سی و دومین کنفرانس بین المللی مهندسی برق

Optimal D2D Resource Allocation in Heterogeneous Cellular Networks by Decentralized Multi-Agent Deep Q-Learning

نویسندگان :

Pouya Akhoundzadeh¹ Ghasem Mirjalily² Mohammad taghi Sadeghi³

1- Yazd university 2- Yazd university 3- Yazd university

کلمات کلیدی :

communication،resource allocation،interference management،multi agent،Deep Q-learning

چکیده :

This paper explores efficient resource allocation method in the Heterogenous cellular networks powered by device-to-device (D2D) communication, which allows mobile users to connect directly, thus enhancing spectrum utilization and system capacity. However, this direct communication can lead to increased interference and potentially degrade system performance. Addressing the challenge of allocating resources effectively in the presence of D2D communication, we propose a decentralized multi-agent deep Q-Learning approach named Autonomous Multi-Agent Deep Q-Learning (AMADQL). This algorithm empowers D2D users to act as individual agents that autonomously adjust their spectrum and power levels, aiming to minimize interference without depending on the network infrastructure. The proposed method is tested through simulations, where it showcases promising convergence and superior performance metrics compared to traditional resource allocation strategies.