Resource Allocation in THz-NOMA-Enabled HAP Systems: A Deep Reinforcement Learning Approach

Aerial access networks have been considered as a key-enabling concept for addressing emerging challenges of future 6G wireless networks. In this paper, we focus on using the higher-altitude tier of aerial access networks, high altitude platform (HAP) as a flying base station at the top of a cellular network. The HAP serves a set of ground users (GUs) that are grouped using a non-orthogonal multiple access technique in a downlink terahertz-based communication system. We aim to augment the achievable transmission rate of GUs by jointly optimizing the power allocation, transmission bandwidth, antenna beamwidth, and altitude of the HAP under limited resources and the users’ minimum data rate requirements. Considering network dynamics and problem complexity, we design an intelligent joint HAP control and resource allocation scheme based on a deep reinforcement learning algorithm. Specifically, a deep deterministic policy gradient algorithm is proposed to find the optimal solution as well as maximize the long-term reward for the considered problem with continuous decision variables. Numerical simulation results show that our proposed scheme outperforms alternative schemes.