User Tools

Site Tools


10a.002.TAU_WP7 - Learning-Aided Mobile THz Communications

Project - Summary

The terahertz (THz, 0.3­3 THz) band offering tens of gigahertz of consecutive bandwidth is nowadays considered as a major candidate for new radio access technology for 6G cellular systems. By utilizing this bandwidth one may not only provide extreme data rates but enable principally new applications such as holographic telepresence and virtual reality. The sub­millimeter wavelength promises ultra large antenna arrays capable of creating extremely directional steerable antenna radiation patterns with the beamwidth of just a few degrees or even less. This feature is vital for THz communications not only allowing to overcome severe path loss at these frequencies but ensuring almost interference free environment. In this project, we plan to develop Machine Learning (ML) ­aided ray­tracing simulation methodology capable of representing dynamically changing propagation conditions in real­time for extension of propagation models obtained for specific environments to other typical deployment options.

Project - Team

Team Member Role Email Phone Number Academic Site/IAB
Evgeny Kucheryavy PI 358 40 771 0619 Tampere University
Roman Kovalchukov Co-PI N/A Tampere University
Alex Pyattev Project Mentor N/A YL-Verkot Oy

Project - Novelty of Approach

There is a number of works aimed at optimizing and accelerating the wireless ray tracing procedure, for example, a visibility graph based method [1, 2]. This method involves dividing the field of view of the transmitter into zones, taking into account line of sight, reflections and diffraction, compiling a visibility graph based on them and searching for zones in which the transmitter is located. The authors in [3] proposed to use Fermat's least principle to convert data from 2D to 3D. In addition to this, the optimal number of reflections that should be taken into account in the simulation was investigated. To simplify ray tracing, a novel method was proposed in [4] for dividing the space into smaller ones to distribute objects over them. In ray tracing, only the areas in which propagation occurs are considered. The authors in [5] proposed to remove from the database, which contain a lot of data on the propagation of radio waves, as well as a geometric method based on smoothing irregularities in buildings with a complex shape. In [6], rays that have undergone the same interaction with surrounding objects are proposed to be combined into single “entities” to reduce the number of rays and, as a result, accelerate ray tracing. The authors in [7] presented a method for compiling a visibility table, in which all reflections are calculated in advance, for subsequent use in modeling. In [8], using machine learning methods, the influence of objects such as buildings, distance, as well as smaller objects on the result of modeling using ray tracing was plotted. Thus, the authors of previous studies used mostly deterministic methods to accelerate ray tracing, with the exception of the latest work, where the authors used machine learning to explore the degree to which external factors influence the result. However, no one has used neural networks to speed up the ray tracing method, so using them to increase the performance of ray tracing in modeling radio channels is relevant.

Project - Deliverables

1 Cluster-based stochastic propagation models
2 Implementation of the developed algorithms in the simulation tool

Project - Benefits to IAB

The developed algorithms and their implementation in ray-tracing simulation software will be especially beneficial for YL-Verkot Oy since it can be used to adequately simulate THx channels, where YL-Verkot Oy accumulates industrial expertise.

Project - Documents

projects/year10/10a.002.tau_wp7.txt · Last modified: 2022/10/10 08:28 by sally.johnson