Generative AI Agent Empowered Multi-User Beamforming Design for HAP Downlink Communications

The high altitude platform (HAP) network has emerged as an essential network component of the emerging sixth-generation of mobile communication systems. This paper investigates the power consumption optimization for HAP downlink communications with the assistance of a designed generative artificial intelligence (AI) framework. The AI architecture incorporates the unique operational characteristics of the HAP network. Assisted by the AI agent, a beamforming optimization problem is formulated to enhance user quality of service (QoS) and improve the energy efficiency (EE) of HAP downlink communications. A QoS-enhanced energy-efficient (Q3E) beamforming algorithm is proposed to solve this problem. The Q3E algorithm employs an artificial neural network architecture without training by supervised datasets to accelerate the solution of the beamforming problem. The simulation results demonstrate that the proposed Q3E algorithm achieves significant performance improvements compared to benchmarks.