Multi-layer RIS on Edge: Communication, Computation and Wireless Power Transfer

TL;DR

This work introduces a multi-layer RIS-based universal paradigm at the network edge to jointly support MIMO communication, edge computation, and wireless power transfer for IoT. It details the RIS structure, inter-layer EM propagation, and wave-domain processing capabilities, and provides prototype demonstrations across MIMO (2×2), D^2NN-like computation with latency below 10 ns, and SWIPT/WPT. The results indicate capacity gains with layering (optimal around three layers), low-latency, power-efficient computation, and enhanced energy transfer, underscoring a potential paradigm shift from digital hardware to all-wave processing for green, scalable edge networks. The paper also outlines practical challenges and future directions, including inter-layer channel calibration, resource scheduling, CSI estimation, and federated edge training, to enable real-world deployment.

Abstract

The rapid expansion of Internet of Things (IoT) and its integration into various applications highlight the need for advanced communication, computation, and energy transfer techniques. However, the traditional hardware-based evolution of communication systems faces challenges due to excessive power consumption and prohibitive hardware cost. With the rapid advancement of reconfigurable intelligent surface (RIS), a new approach by parallel stacking a series of RIS, i.e., multi-layer RIS, has been proposed. Benefiting from the characteristics of scalability, passivity, low cost, and enhanced computation capability, multi-layer RIS is a promising technology for future massive IoT scenarios. Thus, this article proposes a multi-layer RIS-based universal paradigm at the network edge, enabling three functions, i.e., multiple-input multiple-output (MIMO) communication, computation, and wireless power transfer (WPT). Starting by picturing the possible applications of multi-layer RIS, we explore the potential signal transmission links, energy transmission links, and computation processes in IoT scenarios, showing its ability to handle on-edge IoT tasks and associated green challenges. Then, these three key functions are analyzed respectively in detail, showing the advantages of the proposed scheme, compared with the traditional hardware-based scheme. To facilitate the implementation of this new paradigm into reality, we list the dominant future research directions at last, such as inter-layer channel modeling, resource allocation and scheduling, channel estimation, and edge training. It is anticipated that multi-layer RIS will contribute to more energy-efficient wireless networks in the future by introducing a revolutionary paradigm shift to an all-wave-based approach.

Figures (6)

• Figure 1: The structure of multi-layer RIS. The entire structure should be enclosed within a container lined with absorbing materials to minimize energy loss and shield the inter-layer channels from external interference. The size of each meta-atom typically ranges from 1/10 to 1/2 of the wavelength. Meanwhile, the spacing between adjacent RIS layers is generally on the same order of magnitude as the wavelength. The distance between adjacent layers and the size of each meta-atom can be adjusted to meet the requirements of specific tasks, serving as optimizable parameters. Owing to the compact inter-layer spacing and the dense arrangement of passive meta-atoms, multi-layer RIS achieves a miniaturized structure with high scalability.
• Figure 2: Prototypes of RIS-based MIMO transmission, computation and WPT, where, (a) shows two prototypes of RIS-based MIMO transmission. The first is a single-layer RIS prototype designed for point-to-point MIMO video transmission. The second is a multi-layer RIS prototype, which demonstrates enhanced performance compared to the single-layer prototype when the inter-spacing between adjacent layers is less than half the wavelength. (b) shows a three-layer RIS-based computation prototype that successfully performs handwritten digit classification at X-band, achieving a processing latency of less than 10 ns. (c) shows a RIS-based prototype realizing WPT and SWIPT on the right. The dimensions of the prototype are around 0.5m$\times$0.5m$\times$0.04m, close to the size of a regular household appliance.
• Figure 3: System architecture of multi-layer RIS-empowered edge IoT scenario, where multi-layer RIS is leveraged to facilitate the communications, computation, and long-term operation of edge IoT devices. This enables various applications such as smart cities, health, farms, homes, and factories, contributing to a smarter and more connected world.
• Figure 4: System capacity of multi-layer RIS-enabled MIMO transmission with respect to the number of RIS layers.
• Figure 5: Multi-layer RIS-enabled computation schemes, where multi-layer RIS is utilized to perform (a) data-class-specific encryption; (b) DOA estimation; (c) over-the-air computation; (d) joint source-channel coding; (e) image classification.