Skip to Main content Skip to Navigation
Theses

Advanced Linear and Deep Learning based Channel Estimation Techniques in Doubly Dispersive Environments

Abstract : Wireless communications revolution plays a significant role in facilitating several mobile applications like unmanned aerial vehicles, high-speed railway, and vehicular communications. Particularly, the concept of connected vehicles brings a new level of connectivity to vehicles. Along with novel on board computing and sensing technologies, vehicular networks serve as a key enabler of intelligent transportation systems and smart cities. This new generation of networks have a profound impact on the society, making every day traveling safer, greener, and more efficient and comfortable. However, in vehicular environments, the propagation medium between the network nodes is highly time-varying leading to considerable reliability challenges. In fact, transmitted signals propagate through multiple paths, each with a different delay, attenuation, in addition to Doppler shift effect resulting from the motion of vehicles and the surrounding environment. Ensuring communication reliability by the means of accurate channel estimation in such environments is very important. Therefore, the accuracy of the channel estimation influences the system performance, since a precisely estimated channel response influences the follow-up equalization, demodulation, and decoding operations at the receiver. In literature, there exists an extensive work on conventional channel estimation for vehicular communications. However, these conventional estimators rely on many assumptions that limit their performance in highly dynamic time-varying channels. Moreover, linear conventional estimators are impractical solutions in real case scenarios as they rely on statistical models and require high implementation complexity. Although there exists simple linear estimation with affordable complexity, they lack robustness in highly dynamic environments. Therefore, investigating estimators with a good trade-off complexity vs. performance is a significant task. As a prevailing approach to AI, deep learning (DL) develops efficient methods to analyze data by finding patterns and learning underlying structures and represents an effective data driven approach to problems encountered in various scientific fields. The main reason behind integrating DL in wireless communications is to find solutions to communication problems where analytical solutions are intractable or highly complex. DL has a strong ability to overcome this challenge via low-complexity and robust solutions that improve the performance of wireless systems. Additionally, the GPU-based distributed processing enables the DL employment in real-time applications. As a result, DL can be leveraged to exploit the data generated in vehicular networks. In this context, this thesis aims to investigate how to adapt such tools to account for the characteristics of high mobility vehicular networks. We show that integrating optimized DL architectures brings low-complexity solutions for vehicular channel estimation either by improving the performance compared to the simplified linear channel estimators, or by approaching the performance of complex robust model-based estimators with feasible implementation. Therefore, unlike conventional estimators, DL-based estimators provide a good trade-off between the computational complexity and the system performance. Moreover, the generalization ability gives robustness to the system when deployed in highly dynamic environments.
Complete list of metadata

https://hal.archives-ouvertes.fr/tel-03482053
Contributor : Abdul Karim Gizzini Connect in order to contact the contributor
Submitted on : Wednesday, December 15, 2021 - 4:29:19 PM
Last modification on : Sunday, June 26, 2022 - 3:25:43 AM
Long-term archiving on: : Wednesday, March 16, 2022 - 7:27:34 PM

File

20211214_Abdulkarim_GIZZINI_Ph...
Files produced by the author(s)

Identifiers

  • HAL Id : tel-03482053, version 1

Citation

Abdul Karim Gizzini. Advanced Linear and Deep Learning based Channel Estimation Techniques in Doubly Dispersive Environments. Information Theory [cs.IT]. Cergy Paris CY Université, 2021. English. ⟨tel-03482053⟩

Share

Metrics

Record views

93

Files downloads

19