"Joint Communications and Sensing Hybrid Beamforming Design via Deep Unfolding" presented by Dr. Nhan Thanh Nguyen, Based on the IEEE Xplore® article under the same title. Published: IEEE Journal of Selected Topics in Signal Processing, September 2024. Visit article article here.
Joint communications and sensing is expected to be a key capability of future wireless networks, enabling communication and environmental awareness within a unified system. In massive multiple antenna systems, hybrid beamforming offers an effective balance between beamforming performance, hardware cost, and power consumption. However, hybrid beamforming design for joint communications and sensing remains challenging due to the coupling between analog and digital precoders, as well as the need to balance communication throughput with sensing accuracy.
To address these challenges, this webinar presents a fast hybrid beamforming design based on deep unfolding. The featured approach first examines the structure of the joint optimization problem and identifies key differences in the roles of the analog and digital precoders. Based on this insight, a modified projected gradient ascent method is introduced to significantly improve convergence efficiency. This method is then extended using a deep unfolded framework, where learnable parameters further accelerate convergence while fully maintaining the interpretability and flexibility of the original algorithm.
Finally, simulation results are shared, demonstrating notable improvements in both communication performance and sensing accuracy compared with conventional designs. This is achieved alongside a significant reduction in computational complexity and runtime. Ultimately, the webinar demonstrates how model-driven learning can successfully enable efficient and scalable joint communications and sensing solutions.
