Machine Learning-Based Resource Allocation for IRS-Aided UAV Networks

The applications of networking have undergone drastic evolutions since the commercialization of the first generation (1 G) wireless systems. These evolutions have gone from human-to-human (H2H) voice and text communications to the seamless network of people, devices, and objects, which exchange information to assist, accelerate, monitor, and actuate different social, personal, and interpersonal processes. The next-generation systems envision a unified networking architecture which meets the diverse technical specifications of industry 5.0 systems. Currently deployed systems do not provide an eco-friendly architecture for the mass-scale integration of wireless networking. Additionally, the high capital costs of deployment hinder the realization of seamless connectivity for the next industrial revolution. Intelligent reflecting surfaces (IRS) have emerged as a potentially revolutionary and environmentally sustainable technology to enable the concentrated real-world networks required to support advanced cyber-physical systems in the Internet-of-everything (IoE). In this work, we propose a machine learning-enabled resource allocation architecture for IRS-aided unmanned aerial vehicle (UAV) networks. Simulations show that the proposed scheme outperforms traditional resource allocation systems, and is feasible for real-time wireless network optimizations.