Symeon Papavassiliou’s research while affiliated with National Technical University of Athens and other places

Artificial intelligence (AI) has captivated research in network resource orchestration, driving advancements and innovations, including the much-anticipated integration of terrestrial networks (TNs) and non-terrestrial networks (NTNs). Although AI has already been incorporated into standards to complement or replace network functions, its reliability in maintaining the required quality of service (QoS) in highly dynamic and complex environments is often overlooked. This article highlights the significant contributions of AI to unified resource management on the ground and in various orbits while addressing its reliability challenges in complex integrated TN/NTN systems. In this context, the different phases of the AI lifecycle are analyzed and their key processes and steps are identified to introduce mechanisms that ensure reliability at each stage. A principled reliable AI framework is ultimately designed to meet the stringent reliability requirements of highly volatile environments like the integrated TNs/NTNs. To shed light on the practical application of the framework, a comprehensive example is provided for the problem of computation task offloading in integrated TNs/NTNs. This proof of concept emphasizes the design of fully decentralized and scalable solutions while examining the significance and impact of mechanisms that respond to unseen states and maintain the required QoS level safely. Overall, this article provides a guide for tackling the challenges of ensuring reliability in AI-enabled TN/NTN functions that can serve as a powerful catalyst for ongoing standardization efforts.

Wireless Federated Learning (FL) is a distributed Artificial Intelligence (AI) framework, enabling decision-making at the network edge where data are generated. However, wireless transmissions of model updates from edge nodes to the coordinating server are vulnerable to jamming, alongside the inherent risk of poisoning the learning process. In this paper, we tackle the problem of coordinated jamming and poisoning attacks in wireless FL networks, where malicious edge nodes disrupt transmissions of legitimate local model updates to the cloud server while injecting poisoned model updates to manipulate the global model. To this end, we introduce two complementary mechanisms operating alternately. First, a robust global model aggregation algorithm is developed to address poisoning attacks by weighting edge nodes’ local model updates using a novel contribution index. The calculation of the index is inspired by the Shapley value, but it offers polynomial complexity compared to existing methods. Subsequently, a distributed power control solution for jamming attack mitigation in the uplink of the FL network is introduced based on Bayesian games with incomplete information. Both legitimate and malicious nodes aim to successfully transmit their model parameters, minimizing transmission power and time to the server, while having probabilistic knowledge about the malicious behavior of the other nodes in the game. The proposed unified approach and each individual mechanism are assessed via modeling and simulation, verifying their effectiveness in mitigating both attacks while achieving a good tradeoff between global model accuracy and consumed time and energy compared to state-of-the-art approaches.

6G networks envision a seamless integration of data and AI into their core operations, marking a new era where distributed client devices actively engage in intelligent processes, supported by a diverse ecosystem of AI service providers. This shift necessitates approaching traditional resource allocation problems from a market perspective, transparently reflecting the gains and losses of clients and providers within the ecosystem. In this paper, we use a multi-server, multi-model Federated Learning (FL) network as a case study to propose a market-based resource allocation framework. The framework models the joint problem of client resource allocation and pricing as a Fisher market, where servers compete to attract clients with private data and local computing resources for distributed model training. The joint problem is formulated as a convex program, aiming to maximize the total server utility related to their model's accuracy, while incorporating clients' total computing resource constraints and servers' budget constraints. An analytical solution to the convex program is derived concluding the so-called Market Equilibrium (ME) point, where market clearance is achieved. The effectiveness of the proposed market-based resource allocation framework is validated through comparisons with established resource allocation and sharing schemes from the literature.

In the field of autonomous transportation systems, the integration of Unmanned Aerial Vehicles (UAVs) in emergency response scenarios is important for enhancing the operational efficiency and the victims’ positioning. This paper presents a novel Positioning, Navigation, and Timing (PNT) framework, named HEROES, which leverages the UAV and Integrated Sensing and Communication (ISAC) technologies to address the challenges in post-disaster environments. Our approach focuses on a comprehensive post-disaster scenario involving multiple victims, first responders, UAVs, and an Emergency Control Center (ECC). HEROES enables UAVs to function as anchor nodes and facilitate the precise positioning of the victims while simultaneously collecting critical data from the disaster area. We further introduce a Reinforcement Learning (RL) model based on the Optimistic Q-learning with Upper Bound Confidence algorithm, enabling the victims and first responders to autonomously select the most advantageous UAV connections based on their channel gain, shadowing probability, and positional characteristics. Furthermore, HEROES is based on a Satisfaction game-theoretic model to enhance the sensing, communication, and positioning functionalities. Our analysis reveals the existence of various satisfaction equilibria, including Minimum Efficient Satisfaction Equilibrium (MESE), ensuring that the UAVs meet their Quality of Service (QoS) constraints at minimal operational costs. Extensive experimental results validate the scalability and performance of HEROES, demonstrating significant improvements over existing state-of-the-art methods in delivering PNT services during humanitarian emergencies.

The automotive industry has been one of the vertical sectors eagerly waiting for the extended availability of 5G connectivity in order to deliver Connected and Automated Mobility (CAM) services. These services require extremely fast and reliable, uninterrupted communication to guarantee the safety of the drivers and other road users. Even though extended analysis and evaluation of the expected performance of 5G for CAM services has taken place in the past years via simulation studies and local trials based on 5G experimental testbeds, performance evaluation based on real 5G networks has been extremely limited, due to their unavailability until recently. Even more so in ural/highway conditions, as the 5G deployments so far have been focused on urban environments with greater population coverage. This article is among the first to present evaluation data and the corresponding analysis of the 5G Non-Stand Alone (NSA) network performance for CAM services based on neighboring 5G (overlay) networks in the cross-border corridor between Greece and Turkey, by one of the leading global 5G vendors and two of the top national operators. The performance evaluation focuses on the effect of inter-PLMN (Public Land Mobile Network) Handovers on the throughput, latency and interruption time experienced by a mobile user, and the network metrics achievable under various network configurations.

The integration of Knowledge Graphs (KGs) with Question Answering (QA) systems is transforming the landscape of Artificial Intelligence (AI). Through the combination of these technologies, novel features can be provided for the translation of questions in natural language into database queries. Even if a lot of work is emerging in this domain, this is not the case when we refer to translation of text to Cypher queries, where Cypher is one of the dominant query languages used for the development of KGs (e.g., based on the Neo4j technology). In this context, this paper provides a robust and efficient framework to systematically assess the efficiency of Large Language Models (LLMs) to support Text-to-Cypher conversion, focusing on the evaluation of open-source LLMs. The framework utilizes metrics and validators offered by an open-source software library that we developed, called CyVer. This study also assesses the impact of different schema representations of the KG on schema-aware query generation and the performance of LLMs on questions of different complexity requiring a depth of reasoning on the KG. A case study is described based on the application of the detailed framework in a KG with a large and complex schema that hosts data to track information related to the Sustainable Development Goals (SDGs). The experimental results demonstrate the effectiveness of the proposed framework, highlight the importance of the size of open-source models in the semantic comprehension of questions and the generation of valid Cypher queries, and stress the challenge for the generation of accurate queries in the case of questions requiring complex Cypher logic.

The application of Social and Emotional Learning (SEL) methodologies within classrooms has been widely adopted and examined, highlighting its positive impact on the academic, emotional, and social sectors at both the individual and classroom levels. Through this wide applicability, a need is identified to focus more on the development of SEL approaches that support the integration of the various parts of a SEL methodology (e.g., planning, implementation, monitoring, and assessment). Such approaches should not underestimate the value of the SEL assessment. It is necessary to engage both the developers and practitioners of SEL methodologies and provide tools to assist teachers to better interpret the collected data and results and plan targeted SEL activities. Motivated by such needs, we detail the EduCardia SEL methodology that focuses on the provision of a set of guidelines and tools to assess and improve the social and emotional competencies of students in primary and secondary educational levels, by taking advantage of Information and Communication Technologies (ICT). The EduCardia SEL methodology considers the integration of the various parts of a SEL methodological approach and supports continuous assessment processes, while offering user-friendly interfaces to interpret the produced results. Evaluation results are provided based on the application of the EduCardia SEL methodology in a large set of K12 schools across Greece for a short-term period, considering three different age groups. It is shown that the targeted competencies are improved, while insights are provided by teachers for the applicability and efficiency of the proposed approach.