<p>A primary challenge of terahertz communications is the high free-space path loss experienced by THz signals, rendering omnidirectional antennas inadequate for communications needs in this spectral range. One possible solution lies in using types of antennas that emit radiation in specific directions, such as leaky-wave antennas (LWAs), in order to direct power in the exact direction of the receiver. However, for an arbitrary arrangement of receivers, designing an antenna that emits power in all of their directions while avoiding wasting power elsewhere is a significant challenge. In this paper, we propose an irregular, aperiodic design schema for LWAs that can generate emission maxima simultaneously at a variety of angles. We combine this novel antenna architecture with the power of neural networks in order to build a model that can accurately and quickly select a design from among the extremely large set of options, that will produce a desired far-field signal. Combined with the ability to perform experimental tests using rapid-prototyping methods, we propose an end-to-end solution for antenna design that allows researchers to both generate and fabricate a slot design for some chosen far-field signal within minutes, reducing the need for lengthy finite-element calculations or tedious trial-and-error experiments.</p>

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Using Neural Networks to Design Leaky-Wave Antennas for Terahertz Wireless Links

  • Joshua Neronha,
  • Hichem Guerboukha,
  • Daniel M. Mittleman

摘要

A primary challenge of terahertz communications is the high free-space path loss experienced by THz signals, rendering omnidirectional antennas inadequate for communications needs in this spectral range. One possible solution lies in using types of antennas that emit radiation in specific directions, such as leaky-wave antennas (LWAs), in order to direct power in the exact direction of the receiver. However, for an arbitrary arrangement of receivers, designing an antenna that emits power in all of their directions while avoiding wasting power elsewhere is a significant challenge. In this paper, we propose an irregular, aperiodic design schema for LWAs that can generate emission maxima simultaneously at a variety of angles. We combine this novel antenna architecture with the power of neural networks in order to build a model that can accurately and quickly select a design from among the extremely large set of options, that will produce a desired far-field signal. Combined with the ability to perform experimental tests using rapid-prototyping methods, we propose an end-to-end solution for antenna design that allows researchers to both generate and fabricate a slot design for some chosen far-field signal within minutes, reducing the need for lengthy finite-element calculations or tedious trial-and-error experiments.