Aristotle University Of Thessaloniki (AUTH)
Optimization of intelligent wireless semantic communications for 6g and beyond
Abstract
dc:descriptionIn recent years, wireless communications have become a pivotal part of everyday life by enabling many applications. In order to achieve this, low-latency, reliable networks with enhanced capabilities in terms of coverage, connectivity and resource management are required. Due to this, the rise of 6G and beyond networks is necessitated, since conventional communication systems struggle to cope with the vast amount of generated data and the stringent requirements that are imposed by various applications such as smart cities, autonomous driving, distributed learning and virtual reality. In contrast to conventional communication, where perfect raw data recovery is desired, in many of these applications, the main objective is the delivery of meaningful information in order to execute the specific task at hand. To this end, this thesis investigates the use of the underlying semantics for different goals such as communication, image semantic segmentation and computing, by leveraging novel communication techniques and tools such as semantic communications (SemComs), over-the-air (OTA) computing, split learning and deep neural networks (DNNs) in order to establish next generation networks. Initially, we investigate the use of semantics for communication. For the first time in the literature, we propose a hybrid multicarrier framework that utilizes both bit communication (BitCom) and semantic communication (SemCom) for text transmission. In particular, the considered system jointly selects the transmission mode of each subcarrier and optimizes power allocation to minimize two significant delay-related metrics, namely sum delay and maximum delay of all subcarriers, while satisfying strict similarity constraints. To tackle the non-convexity of the resulting problems alternating optimization and an optimal iterative heuristic algorithm with low complexity are utilized for minimization. Apart from these, a novel association of data to subcarriers is also introduced based on the channel gain of each subcarrier, ultimately further enhancing the performance of the proposed hybrid system. Based on the simulation results, it is observed that while SemCom improves the performance for a wide range of signal-to-noise ratio (SNR), it is not always the preferable transmission mode, thus illustrating the advantage of the proposed hybrid framework in order to establish low-latency communication. Aiming to extend the introduced hybrid BitCom-SemCom system into a multi-user network scenario, we then focus on the performance of a hybrid cellular network, where each user selects the optimal transmission mode. In this direction, we derive useful analytical expressions for the network outage probability. In addition to this, we propose a generalized version of this metric, which examines the probability of a specific range of users to be in outage. Also in an attempt to consider the practical limitations of SemComs related to the increased memory and computational requirements, we introduce another metric, called semantic utilization. Based on the analysis of the aforementioned metrics, useful insights for the performance and the design of such systems through cell radius optimization are extracted, showcasing that the studied system can support massive connectivity, while achieving enhanced network reliability and coverage. Moving from the hybrid collaboration of BitComs and SemComs for text transmission, we delve into enabling another capability related with artificial intelligence that can be pivotal for various 6G and beyond applications, namely image semantic segmentation. To this end, we illustrate a novel split learning-based method that can be leveraged to minimize the overall inference time of the network. Moreover, to ensure the practicality of our technique, two transmission schemes are investigated and various resource allocation policies are studied in order to provide near-optimal performance with significantly reduced complexity. In this way, the proposed system can achieve low latency communication, while efficiently managing resources, hence making it suitable for next generation networks. After this, we study OTA computing in order to establish the use of semantics for computing, which is another critical application associated with 6G and beyond networks. With this in mind, waveform design is investigated for narrowband and wideband channels under practical impairments related to time and frequency synchronization. For the narrowband channel case, a novel DNN framework is proposed in order to generate new waveforms that utilize the same spectrum, while improving performance under intersymbol interference caused by time synchronization errors. For the wideband channel case, a novel orthogonal frequency division multiplexing-based technique is considered under intercarrier interference caused by frequency synchronization errors. To optimize the performance of OTA computing, a thorough theoretical analysis is conducted and optimal power allocation policies are derived through alternating optimization. As such, by leveraging signal superposition to efficiently utilize communication and computational resources, OTA computing is established as a viable technique for wireless aggregation in massively connected networks. Finally, the conclusions of the current thesis are presented and some possible future directions are briefly discussed.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Evgenidis, Nikolaos
- Ευγενίδης, Νικόλαος
Subjects
dc:subject × 13- Ασύρματες επικοινωνίες έκτης γενιάς
- Σημασιολογικές επικοινωνίες
- Νευρωνικά δίκτυα
- Wireless communications
- Semantic communications
- Deep neural networks
- Over-the-air computation
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Επιστήμες Μηχανικού και Τεχνολογία
- Μηχανική και συστήματα επικοινωνιών, Τηλεπικοινωνίες
- Electrical Engineering, Electronic Engineering, Information Engineering
- Engineering and Technology
- Communication engineering and systems, Telecommunications
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/62875