All-Dielectric Frequency-Diverse Terahertz Antenna for Multi-Source Direction-of-Arrival Estimation

Abstract The sub-terahertz frequency region is a promising candidate for realizing 6Gbeyond mobile networks. However, sub-terahertz waves are subjected to strong free-path loss, requiring highly directional antennas. Precise direction-of-arrival (DoA) estimation is therefore critical to maintain accurate alignment between transceivers. This paper proposes an all-dielectric frequency-diverse antenna with strongly frequency-dependent radiation patterns, which yield directiondependent received spectra for DoA estimation. The antenna operates within the IEEE 802.15.3d band and is optimized for improved manufacturability and reproducibility. Beyond single-source detection, the approach is extended to multi-source DoA estimation. A neural network model is trained on the antenna’s radiation patterns and evaluated for detecting both single-source and multi-source scenarios. Experimental results indicate that DoA tracking can be achieved for transmitter-receiver distances of up to 900 mm with an average RMSE of 2.53 ◦ for single source detection and 4.56 ◦ for multiple sources. These results validate the effectiveness of the proposed system as a robust, datadriven solution for DoA estimation in next-generation terahertz communications systems.