The path towards ultra-reliable low-latency communications via HARQ

Battistella Nadas, Joao Pedro (2021) The path towards ultra-reliable low-latency communications via HARQ. PhD thesis, University of Glasgow.

Abstract

Ultra-reliable Low-latency Communications (URLLC) is potentially one of the most disruptive communication paradigms offered by the next generation of wireless networks, 5G. This is easily demonstrated by the diverse set of applications it enables, such as autonomous driving; remote surgery; wireless networked control systems; mission-critical machine type communication; and many more. Basically, URLLC consists of the almost 100% guarantee of message delivery within a very short time interval. Furthermore, the pressure from climate change coupled with the massive growth of cellular networks expected to occur in the near future means that URLLC must also be energy efficient. On its own, achieving low-latency with high reliability is already a stringent requirement, but when that is coupled with the need for resource efficiency, it becomes even more challenging. That is the motivation behind this thesis: to study URLLC in the context of resource efficiency. Thus, a study of the counterintuitive use of retransmissions, more specifically Hybrid Automatic Repeat Request (HARQ), in the scenario of URLLC is proposed and carried out. HARQ is very attractive in terms of resource efficiency, and that is the motivation behind using it even when stringent time constraints are imposed. Four contributions are made by the present work. Firstly, a mathematical problem is presented and solved for optimizing the number of allowed retransmission rounds considering HARQ in URLLC, considering both energy efficiency as well as electromagnetic irradiation. This representation relies on a few assumptions in order to be realizable in practical scenarios. Namely, these assumptions are regarding the possibility of early error detection for sending the feedback signals and on not having to consider medium access control introduced delays. Secondly, we consider one important aspect of wireless systems, which is that they can be greatly optimized if they are designed with a specific application in mind. Based on this, a study of the use of HARQ specifically tuned for Networked Control Systems is presented, taking into account the particular characteristics of these applications. Results here show that fine-tuning for the specific characteristics of these applications yields better results when compared to using the results from the previous contribution, which are more application-agnostic. These improved results are possible thanks to the exploitation of application-specific characteristics, more specifically the use of a packetized predictive control strategy jointly designed with the communication protocol. Next, the concept of HARQ for URLLC is extended to a larger scale in an effort to relax the aforementioned assumptions. This is studied within the framework of self-organizing networks and leverages machine learning algorithms in order to overcome those strict assumptions from the first contribution. This is demonstrated by developing a digital twin simulation of the city of Glasgow and generating a large dataset of users in the cellular network, which is a third contribution of this thesis. Then, machine learning (more specifically long short-term convolutional neural networks) is applied for predicting message failures. Lastly, a protocol to exploit such predictions in combination with HARQ to deliver downlink URLLC is applied, resulting in a fourth contribution. In summary, this thesis presents a latency aware HARQ technique which is shown to be very efficient. We show that it uses up as much as 18 times less energy than a frequency diversity strategy and that it can emit more than 10 times less energy electromagnetic field radiation when compared to the same strategy. We also propose joint design techniques, where communication and control parameters are tweaked at the same time, enabling wireless control systems with a three-fold reduction in required bandwidth to achieve URLLC requirements. Lastly, we present a digital twin of the city of Glasgow which enables us to create a prediction algorithm for predicting channel quality with very high accuracy—root mean square error on the order of 10−2. This ties into the rest of the contributions as it can be used to enable early feedback detection, which in turn can be used to make sure the latency aware protocol can be employed.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Subjects: T Technology > T Technology (General)
• Colleges/Schools: College of Science and Engineering > School of Engineering > Systems Power and Energy
• Supervisor's Name: Imran, Professor Muhammad
• Date of Award: 2021
• Depositing User: Theses Team
• Unique ID: glathesis:2021-82293
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 01 Jul 2021 07:58
• Last Modified: 01 Jul 2021 07:58
• Thesis DOI: 10.5525/gla.thesis.82293
• URI: https://theses.gla.ac.uk/id/eprint/82293

Related URLs

• Article DOI
• Article DOI
• Article DOI