Leveraging Actor‐Critic Learning for Joint Transmit Antenna Selection and Resource Allocation in mm‐Wave
D2D
Networks
Pouya Akhoundzadeh, Ghasem Mirjalily ABSTRACT
In this paper, we introduce an innovative autonomous multi‐agent actor‐critic (AMAAC) learning approach for efficient resource allocation in millimeter‐wave (mm‐wave) device‐to‐device (D2D) communication networks. This enables dynamic and real‐time decision‐making, adapting to changing network conditions for optimal resource allocation. By incorporating directional microstrip antennas on the user side, our approach improves the signal concentration, coverage, and connectivity in mm‐wave environments. The objective is joint optimization of the transmit antenna selection (TAS), spectrum assignment (SA), and power level selection (PLS), while ensuring that all users meet the QoS requirements. This approach enhances performance through better adaptability, efficient interference management, and scalability for handling many users. Our extensive simulations demonstrate that the AMAAC algorithm significantly improves both system sum rate and energy efficiency, compared to the related approaches. The results of this research highlight the transformative potential of deep reinforcement learning (DRL) in advancing mm‐wave technologies and underscore the importance of innovative directional antenna designs for efficient wireless networks, addressing the increasing demand for high‐capacity and low‐latency communication systems.