Research

Program: «Ερευνώ – Καινοτομώ» - ΕΣΠΑ 2021–2027

The Smart Beyond project focuses on the development of advanced Internet of Things (IoT) infrastructures for smart cities by integrating sensor networks, big data platforms, and artificial intelligence techniques. The project aims to design optimal strategies for the placement of IoT sensors, improve secure communication between devices, and enable intelligent analysis of urban data through machine learning models. By leveraging real-time environmental, mobility, and infrastructure data, the system supports predictive analytics and data-driven decision-making for urban management. The project contributes to improving energy efficiency, environmental monitoring, and public safety while promoting scalable and interoperable smart city solutions.

ENFIELD (European Lighthouse to Manifest Trustworthy and Green AI)

AI4WaterGuard stands for AI-Driven Detection of Illegal Water Abstractions Using Earth Observation, and it is developed under the ENFIELD IDS.1 challenge, which focuses on detecting potential illegal water abstractions using Artificial Intelligence and Earth Observation. The motivation behind this project is the growing pressure on water resources and the increasing difficulty of identifying non-authorised water withdrawals at scale using conventional monitoring approaches alone. In many regions, in-situ controls are fragmented, costly, and cannot provide continuous spatial coverage. AI4WaterGuard proposes an innovative methodological framework that combines Earth Observation data with Artificial Intelligence in order to identify potentially unauthorised water abstractions in a scalable and evidence-based manner. 

Post-Doctoral Programme

The high availability and performance offered by Cloud computing platforms have enabled and popularized an array of data-driven applications, ranging from serving dynamic website content to performing machine learning and deep analytics. Despite the typical high-speed internet connections between applications and Cloud storage, there is still a huge performance gap compared to accessing data from direct-attached memory or even locally-attached disks. Caching is a technique frequently used for reducing application latency and improving throughput by storing a useful subset of the data in fast access hardware. However, applications must either implement their own custom caching mechanisms or use existing caching systems. While Cloud-based caching solutions such as Amazon ElastiCache do exist, they do not cache the data along the I/O path between the Cloud storage and the application. Rather, the data is cached on dedicated servers and still accessed over the network. Hence, what is currently missing is a caching service that provides seamless access to all important data stored on Cloud storage while taking advantage of locally-attached storage tiers on compute clusters. Our research project aims at designing, implementing, and validating such a service while using machine-learning driven policies for fully automating all relevant data management decisions.

Erasmus+ Programme

The COVID pandemic has triggered drastic changes in the delivery of education and knowledge transfer at global scale. Two dramatic changes have been observed so far. Firstly, the education sector was transformed with the deployment of e-learning and blended learning approaches. Secondly, businesses shifted their work practices using remote work practices. Both phenomena have been widely observed and research in the literature for decades, while there are numerous case studies discussing how universities and organisations use technology to overcome barriers associated with the lack of face-to-face interaction. Therefore, once the pandemic arrived, there were several perspectives arguing that both Universities and Organisations should demonstrate a relatively high level of readiness for a seamless transition towards full deployment of e-learning and e-work. After more than six months of the new ways of studying and working, it is observed that there are significant problems with regards to the effectiveness of the adopted practices. The primary source of these problems is that the study and work cannot be supported by a blended mode, where some (even minimal) face to face contact is possible. In blended learning and working modes, the face-to-face contact serves as a support mechanism for (i) enhanced communication, (ii) effective collaboration and (iii) efficient coordination.

As a result of the pandemic Universities already report a number of issues associated with the transition to a fully online delivery. The impact for knowledge transfer projects is also evident in organisations, as remote work is also affected by the lack of face-to-face meetings. The consortium proposes the creation of a Sharing my Learning (Platform-Network-Toolkit) to support the transition towards e-study and e-work. The focus of the project is on University study, while the project ideas will be also tested at a small-scale pilot for organisational knowledge transfer scenarios.

Improvements in memory, storage devices, and network technologies are constantly exploited by distributed systems in order to meet the increasing data storage and I/O demands of modern large-scale data analytics. We present a novel distributed file system that is aware of storage media (e.g., memory, SSDs, HDDs, NAS) with different capacities and performance characteristics. The system offers a spectrum of usage patterns ranging from fully automating data management to providing explicit control by exposing the storage tiers to users.

An extensive exploration and development of high-throughput multidimensional on-chip interconnection network topologies and architectures using in-silicon photonic interconnects, which are an innovative technology in the field of hybrid photon-electric integrated interconnected Networks-on-Chips. This new technology sparks the invention of more elaborate topologies and network flow control protocols, where their combination has goal to further achieve performance in hybrid integrated interconnected Networks-on-Chips architectures.

The research work concentrated on the creation of hotspots prediction and detection mechanism, where there is an increasing and focused flow of data in the topology of a NoC, and consequently congestion in the data traffic, which dramatically reduces network throughput. Given the need to address the phenomenon as a next step, these hotspots are identified and reduced using a new proposed pro-active and adaptive high-performance routing algorithm.