Research on AI-Assisted Design of Microwave Components using Neural Networks

With the increasing demand for high-performance and advanced multi-functional microwave components, modern designs are becoming significantly more complex, increasingly incorporating emerging technologies such as 3D printing, reconfigurable intelligent design, miniaturization, and so on. These advancements lead to highly nonlinear electromagnetic (EM) responses relative to physical variations, posing substantial challenges to traditional design methodologies. To address these issues, AI-assisted methods such as artificial neural networks (ANNs), deep neural networks (DNNs), and convolutional neural networks (CNNs) have emerged as powerful tools for the parametric modeling of microwave components. Compared to traditional design methods, which suffer from prohibitive computational costs due to repetitive full-wave simulations, NNs offer exceptional learning and generalization capabilities. Once properly trained, NNs can accurately map the complex nonlinear relationships between physical parameters and EM behavior, significantly accelerating the design process and enhancing modeling efficiency.

Reconfigurable Distributed Antennas and Reflecting Surface (RDARS)-Enhanced Communication and Sensing Systems

Reconfigurable distributed antennas and reflecting surface (RDARS) has been recently proposed as a promising architecture. The working mode of each RDARS element can be dynamically adjusted between the connection mode, reflection mode, and absorption mode via the switching network. In particular, the dynamic working mode configuration for the RDARS-aided system brings an extra selection gain compared to the existing reconfigurable intelligent surface (RIS)-aided system and distributed antenna system (DAS). Regarding a RDARS-aided system, several fundamental issues remain unsolved. First, how to design a novel codebook-based beamforming to unleash the potential of RDARS? Second, how to obtain the seamless blend of the optimal beamforming design and RDARS element configurations with low complexity? Finally, how to balance the communication and sensing performance in ISAC systems? In this talk, I will introduce our recent progress on the RDARS architecture, providing efficient solutions to these challenges through joint beamforming and mode selection design.

Dual-Scale Channel Estimation in ISAC Networks

Existing sensing-assisted communication studies always focus on beam tracking and prediction, while the impact of sensing accuracy on the communication rate is not yet well quantified. In this paper, we propose a novel integrated sensing and communication (ISAC)-enabled dual-scale channel estimation framework, where large-scale channel estimation benefits from sensing, and the temporal variation of small-scale channel state information is modeled via channel aging. By characterizing the impact of angular sensing error on the communication spatial correlation matrix, we derive a closed-form expression for the achievable rate under dual-scale channel estimation errors. Considering the different characteristics in time scales, we design the sensing duration for slow-varying large-scale channel and determine the update timing and frequency for fast-varying small-scale channel information within a given frame structure. We formulate an average achievable rate maximization problem under limited time resources and sensing Cramer-Rao bound constraints, and propose a segmented golden-section based joint optimization algorithm to efficiently solve this non-convex problem. Simulation results validate that the system can leverage additional sensing capabilities to enhance communication efficiency.

Joint Active and Passive Sensing Beamforming Design for Multi-UAV Cooperative Integrated Sensing, Communication and Computing

This presentation introduces a multi-UAV cooperative integrated sensing and communication (ISAC) system with mobile edge computing (MEC), addressing the critical challenges of beamforming design and resource optimization in emerging ISAC systems. In the proposed framework, UAVs cooperate to perform joint active and passive sensing of multiple targets while providing MEC services for edge users. To investigate the problem, we formulate a joint optimization of UAV deployment, beamforming, computing resource allocation, and user association, with the objective of minimizing the system energy consumption while guaranteeing the requirements on task processing delay and sensing performance. To address this mixed-integer non-convex optimization problem, we introduce a double layer alternating optimization scheme that partitions the problem into an outer layer alternating optimization and an inner layer alternating optimization. The results validate the superiority of this framework over the baseline algorithms in reducing energy consumption and enhancing sensing performance.