A recent study published in Engineering introduces an innovative design paradigm for reconfigurable intelligent surfaces (RISs), aiming to enhance their design efficiency and versatility. This research, led by a team of scientists from Southeast University and Guangzhou University, offers a fresh approach to address the challenges in RIS design.
RISs, enabled by digital coding technology, are crucial for manipulating electromagnetic (EM) waves in real-time. They play a significant role in 5G and 6G research. However, traditional automatic RIS design methods face issues. Most involve extensive numerical simulations of units, including passive patterns and active devices, leading to high data acquisition and training costs. The commonly used random pixelated method for passive patterns also has drawbacks, such as massive pixel combinations and blocked excitation current flow, which affect design efficiency and effectiveness.
To overcome these problems, the research team proposed a new paradigm that combines a novel topological representation method and a separate design architecture. The non-uniform rational B-spline (NURBS) is introduced to represent continuous patterns. This method can map complex 100-dimensional continuous patterns to five-dimensional NURBS control points, reducing the pattern solution space by 20 times. It ensures smooth excitation current flow, enhancing the resonance ability of surface patterns.
The separate design architecture, based on multiport network theory, divides the RIS unit into four subparts: active device, pattern layer, dielectric layer, and metal ground. This separation simplifies the design process. By using a pre-incremental learning network (PILN) and theoretical calculations, the multistate responses of units with different subpart combinations can be obtained quickly, in nearly 1 second. This architecture reduces dataset acquisition costs by 62.5% and enables dataset and model reuse for different RIS designs.
The researchers validated the new paradigm through three design examples: two typical high-performance RISs and an ultra-wideband multilayer RIS. These designs achieved excellent performance, even outperforming manually crafted units. For instance, a 1-bit phase-modulation RIS unit had an amplitude loss of less than 3 dB within 9~15 GHz and a 50% relative bandwidth.
This new design paradigm provides an efficient solution for RIS automatic design. It paves the way for the development of multifunctional and multi-structure RISs, which are expected to have broad applications in wireless communication, sensing, and other fields. However, the team also acknowledges that there are areas for further improvement, such as incorporating non-continuous patterns and establishing more general rules for pattern domain representation. Future research will focus on overcoming these challenges to further optimize RIS design.
The paper "A High-Efficiency and Versatile Reconfigurable Intelligent Surface Design Paradigm with Novel Topological Representation," authored by Ying Juan Lu, Jia Nan Zhang, Yi Han Zhao, Jun Wei Zhang, Zhen Zhang, Rui Zhe Jiang, Jing Cheng Liang, Hui Dong Li, Jun Yan Dai, Tie Jun Cui, Qiang Cheng. Full text of the open access paper: https://doi.org/10.1016/j.eng.2024.11.028
