.png){kind=icon}
.png){kind=icon}
Return
Language-controllable programmable metasurface empowered by large language models
.png){kind=icon}
OA
AI
DOI：10.1515/nanoph-2023-0646.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Programmable metasurface has become a prominent tool in various areas including control, communication, computing, and so on, due to its unique capability in the electromagnetic (EM) manipulation. However, it is lack of the intelligence in the sense that it usually requires the manual intervention, and thus makes it hard to behavior as the human process. To endow the programmable metasurface with the intelligence, we here proposed the concept of the language-controllable programmable metasurface for autonomous EM manipulations by exploring the notable capability of large language models (LLMs) in attaining the human-like intelligence. We have established a proof-of-principle system of language-controllable programmable metasurface, where, for illustration, the programmable metasurface is designed to have 32 x 24 binary electronically controllable meta-atoms and work at around 5.5 GHz. In addition, we have constructed a visual-semantic map to facilitate the language-controllable EM manipulation in three-dimensional (3D) physical environments. We have experimentally demonstrated that our language-controllable programmable metasurface is capable of decomposing autonomously an ambiguous task of EM manipulation into a sequence of executable ones and implementing them individually in real-world indoor settings. We expect that the presented strategy could hold promising potential in pushing programmable metasurfaces towards human-level autonomous agents, which are capable of accomplishing the smart EM-involved multi-modality manipulations through self-directed planning and actions.
Keywords:
programmable metasurface
large language models
EM manipulation
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
6.6
Papers:
2.9K
Citations:
1.6W