Deep Learning‐Driven Wideband Gain Metasurface for Intelligent Beam Amplification and Tracking

ABSTRACT The rapid advancement of the low‐altitude economy has generated growing demand for intelligent electromagnetic infrastructure, where metasurfaces have emerged as a pivotal platform for controlling electromagnetic waves. Nevertheless, path loss and shadow fading continue to pose significant challenges, substantially constraining performance across diverse application scenarios. Here, we demonstrate an intelligent wideband gain metasurface that enables adaptive operation in changing electromagnetic environments for signal enhancement and dynamic target tracking, assisted by an integrated optoelectronic sensing system. Operating across 3.25–4.75 GHz with a 37.5% relative bandwidth, our intelligent metasurface delivers over 11 dB of stable gain while maintaining excellent phase modulation capability. To efficiently configure the gain metasurface, we develop an inverse design framework that achieves diverse adaptive functions with minimal training samples, enabling precise wavefront control. We fabricated a 4 × 8 metasurface prototype and experimentally validated its exceptional capabilities in electromagnetic amplification and manipulation through a representative application scenario involving tracking a moving unmanned aerial vehicle (UAV). This work paves the way for integrating intelligent metasurfaces into electromagnetic infrastructure and supports their potential application in the low‐altitude economy.