FlexSARCloak: A Flexible SAR Cloak Driven by Task-Oriented Learning.

Invisibility─the remarkable ability to render objects imperceptible─has long been a persistent dream of humankind. However, traditional cloaking materials are typically rigid and inflexible, limiting their adaptability to various shapes and requirements. Even when flexibility is achieved, uncontrollable scattering in complex electromagnetic environments continues to pose significant challenges in the design of flexible cloaks. In this paper, we present a task-oriented learning (TOL)-based metasurface design approach, successfully realizing a broadband, polarization-independent, flexible SAR cloak (FlexSARCloak). Experimental results reveal that the cloak achieves over 90% absorption efficiency across the 9-25 GHz frequency range for incidence angles ranging from 0 to 60°. Outdoor experiments further confirm that the FlexSARCloak achieves near-perfect invisibility under UAV-mounted SAR. Furthermore, a comprehensive series of environmental tests, including exposure to extreme temperatures, variable humidity, and sandstorm conditions, was conducted to simulate natural or extreme conditions, which successfully validated the stability of the electromagnetic performance. The proposed FlexSARCloak presents considerable potential for applications in electromagnetic protection and related domains. Moreover, the introduction of the TOL algorithm is expected to significantly accelerate the design and optimization of metasurfaces, paving the way for the physical realization of multifunctional metasurfaces.