Deep learning-based hybrid beamformer design for millimeter-wave integrated sensing and communication systems

Integrated sensing and communication (ISAC) systems have emerged as a key enabler for next-generation wireless networks, yet the joint optimization of communication and sensing remains a fundamental challenge. In this work, we propose a deep learning-based (DL-based) hybrid beamforming framework that adaptively balances communication sum rate and sensing accuracy for ISAC systems. An optimization strategy is established to dynamically perform resource allocation and joint optimization across both tasks. To further enhance efficiency, we develop a DL-based hybrid beamforming architecture to ensure user fairness in communications while achieving high sensing accuracy. Simulation results demonstrate that the proposed approach achieves a high accuracy in beam synthesis, improved sidelobe suppression, and enhanced trade-offs between communication and sensing. In contrast to conventional optimization-based beamforming methods, the proposed method achieves these gains with lower computational complexity and provides adaptability to dynamic environments.