Over-the-Air Goal-Oriented Communications

2025-12-23

AI Summary

Edge intelligence inference faces stringent constraints on energy, latency, and hardware resources, especially at wireless edge devices. Method: This paper proposes a novel “channel-as-computation” paradigm for over-the-air intelligent inference, leveraging reconfigurable intelligent surfaces (RIS) to perform feature fusion and analog-domain inference directly in the wireless channel—bypassing raw-data reconstruction at the receiver. We introduce the first differentiable physical-layer RIS channel model, integrated end-to-end into deep neural networks as the Metasurfaces-Integrated Neural Network (MINN). MINN employs stacked RIS architectures, task-oriented feature encoding, and physics-aware backpropagation for joint optimization. Contribution/Results: Without compromising inference accuracy, MINN significantly reduces computational overhead at both transmitter and receiver and lowers transmit power. Experiments across multiple benchmarks demonstrate that MINN achieves performance on par with fully digital DNNs while drastically cutting total system energy consumption—enabling native physical-layer co-design of communication and computation.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Search and Optimization: Learning to Search
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Goal-oriented communications offer an attractive alternative to the Shannon-based communication paradigm, where the data is never reconstructed at the Receiver (RX) side. Rather, focusing on the case of edge inference, the Transmitter (TX) and the RX cooperate to exchange features of the input data that will be used to predict an unseen attribute of them, leveraging information from collected data sets. This chapter demonstrates that the wireless channel can be used to perform computations over the data, when equipped with programmable metasurfaces. The end-to-end system of the TX, RX, and MS-based channel is treated as a single deep neural network which is trained through backpropagation to perform inference on unseen data. Using Stacked Intelligent Metasurfaces (SIM), it is shown that this Metasurfaces-Integrated Neural Network (MINN) can achieve performance comparable to fully digital neural networks under various system parameters and data sets. By offloading computations onto the channel itself, important benefits may be achieved in terms of energy consumption, arising from reduced computations at the transceivers and smaller transmission power required for successful inference.

Problem

• Goal-oriented communications for edge inference without data reconstruction
• Wireless channel computation using programmable metasurfaces for neural networks
• Offloading computations to channel reduces energy consumption in inference

Innovation

• Using programmable metasurfaces for wireless channel computations
• Training end-to-end system as single deep neural network
• Offloading computations to channel reduces energy consumption