An open invisible space enabled by reconfigurable metasurfaces and self-play reinforcement learning

Abstract

An open, dynamic, and electromagnetically invisible space has been constructed using reconfigurable metasurfaces and self-play reinforcement learning. A model named MetaSeeker is proposed to optimize the cloaking performance of randomly distributed metasurfaces. The hidden objects can move freely within the constructed invisible space, with environmental similarity of 99.5%. This advancement provides an innovative solution for cloaking technologies in complex environments.

An open invisible space enabled by reconfigurable metasurfaces and self-play reinforcement learning.