AirCNN via Reconfigurable Intelligent Surfaces: Architecture Design and Implementation

2025-10-29

AI Summary

To address the high power consumption, large latency, and limited flexibility of conventional CNN hardware implementations, this paper proposes AirCNN—the first reconfigurable intelligent surface (RIS)-based over-the-air (OTA) analog computing CNN architecture. AirCNN models the wireless channel as a programmable physical-layer neural network, enabling native OTA execution of Conv2d and lightweight ConvSD operations via joint MISO/MIMO transmission, transmit precoding, and receive beamforming—controlled through RIS phase tuning. Its key innovation lies in leveraging the propagation environment itself as a neural computing substrate, supporting multi-RIS coordination and diverse convolutional structures. Simulation results demonstrate that AirCNN achieves competitive accuracy on image classification tasks; the MISO configuration outperforms MIMO in most scenarios; and multi-RIS deployment significantly enhances performance—particularly in line-of-sight-dominant environments.

Technology Category

• Machine Learning: Hardware-aware ML
• Computer Vision: Learning & Optimization for CV
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Security and Privacy: Security and privacy of machine learning and AI applications

Abstract

This paper introduces AirCNN, a novel paradigm for implementing convolutional neural networks (CNNs) via over-the-air (OTA) analog computation. By leveraging multiple reconfigurable intelligent surfaces (RISs) and transceiver designs, we engineer the ambient wireless propagation environment to emulate the operations of a CNN layer. To comprehensively evaluate AirCNN, we consider two types of CNNs, namely classic two-dimensional (2D) convolution (Conv2d) and light-weight convolution, i.e., depthwise separable convolution (ConvSD). For Conv2d realization via OTA computation, we propose and analyze two RIS-aided transmission architectures: multiple-input multiple-output (MIMO) and multiple-input single-output (MISO), balancing transmission overhead and emulation performance. We jointly optimize all parameters, including the transmitter precoder, receiver combiner, and RIS phase shifts, under practical constraints such as transmit power budget and unit-modulus phase shift requirements. We further extend the framework to ConvSD, which requires distinct transmission strategies for depthwise and pointwise convolutions. Simulation results demonstrate that the proposed AirCNN architectures can achieve satisfactory classification performance. Notably, Conv2d MISO consistently outperforms Conv2d MIMO across various settings, while for ConvSD, MISO is superior only under poor channel conditions. Moreover, employing multiple RISs significantly enhances performance compared to a single RIS, especially in line-of-sight (LoS)-dominated wireless environments.

Problem

• Implementing CNNs through wireless analog computation using RIS
• Optimizing RIS architectures for different convolution types and constraints
• Evaluating performance of multi-RIS systems in various channel conditions

Innovation

• Leveraging RIS for wireless CNN emulation
• Optimizing transmitter receiver and RIS parameters
• Extending framework to depthwise separable convolution