Computer Science, Computer Science Applications, Information Systems, Software
6
Scopus Publications
Scopus Publications
Exploring the performance of real-time data imputation to enhance fault tolerance on the edge: A study on environmental data Dimitris Gkoulis, Anargyros Tsadimas, George Kousiouris, Cleopatra Bardaki, Mara Nikolaidou Simulation Modelling Practice and Theory, 2025 Real-time data streams from edge-based IoT sensors are frequently affected by transmission errors, sensor faults, and network disruptions, leading to missing or incomplete data. This paper investigates the application of lightweight, real-time imputation methods to enhance fault tolerance in edge computing systems. To this end, we propose to integrate a modular imputation engine on edge system supporting lightweight forecasting models selected for their computational efficiency and suitability to operate on real-time data streams. To assess the performance of different popular lightweight forecasting models for real-time applications, a simulation framework is introduced that simulates the operation of the imputation engine, replicates sensor failure scenarios and allows controlled testing on real-world systems. Imputation accuracy is evaluated using Mean Absolute Error (MAE), 95th percentile error, and maximum error, with results benchmarked against sensor tolerance thresholds. The simulation framework is used to explore imputation on environmental data based on observations collected from a weather station. The findings show that Holt–Winters Exponential Smoothing delivers the highest accuracy for real-time imputation across environmental variables, outperforming simpler models suited only for short-term gaps. Errors grow with longer forecasts, confirming imputation as a temporary solution. Evaluations against sensor-specific thresholds offer practical insights, and execution profiling proves these models are lightweight enough for deployment on low-power edge devices, enabling real-time, fault-tolerant monitoring without cloud dependence.
Creating interpretable synthetic time series for enhancing the design and implementation of Internet of Things (IoT) solutions Dimitris Gkoulis Internet of Things the Netherlands, 2025 This study establishes a foundation for addressing the challenge of developing Internet of Things (IoT) solutions in the absence of real-world data, a common obstacle in the early stages of IoT design, prototyping, and testing. Motivated by the need for reliable and interpretable synthetic data, this work introduces a structured approach and a dedicated library for creating realistic time series data. The methodology emphasizes flexibility and modularity, allowing for the combination of distinct components–such as trends, seasonality, and noise–to create synthetic data that accurately reflects real-world phenomena while maintaining interpretability. The approach’s utility is demonstrated by creating synthetic air temperature time series, which are rigorously compared against real-world datasets to assess their fidelity. The results validate the proposed methodology’s and library’s effectiveness in producing data that closely mirrors real-world patterns, providing a robust tool for IoT development in data-constrained environments.
Assessing Event Fabrication Methods for Missing Events in Complex Event-Driven IoT Systems: A Smart Farming Case Study Dimitris Gkoulis, Anargyros Tsadimas, Cleopatra Bardaki, George Kousiouris, Mara Nikolaidou International Conference on System of Systems Engineering Sose, 2025 In event-driven IoT systems, real-time data is crucial for effective complex event processing (CEP) and automated decision-making. However, missing sensor events due to network disruptions, sensor failures, or inconsistencies can degrade system performance and analytical accuracy. This study explores event fabrication techniques to reconstruct missing data in an IoTenabled smart farming scenario. We evaluate two methods - Naive and Exponential Smoothing with Linear Trend (ESLT) - by analyzing their accuracy and precision under various nonideal conditions. Experimental results highlight the trade-offs between these approaches, demonstrating that the Naive method, representing memoryless interpolation techniques, is more reliable for short-term gaps and sparse data, while ESLT, as a trend-aware statistical method, performs better with sufficient historical data but struggles in temporally sparse contexts. This distinction serves as a key finding, illustrating fundamental differences in event fabrication strategies. The findings provide insights into optimizing event fabrication strategies for real-time IoT applications, ensuring the continuity in event-driven architectures. Future research will explore adaptive methodologies for dynamically selecting the most suitable fabrication technique based on real-time conditions proposed by this study.
Transforming IoT Events to Meaningful Business Events on the Edge: Implementation for Smart Farming Application Dimitris Gkoulis, Cleopatra Bardaki, George Kousiouris, Mara Nikolaidou Future Internet, 2023 This paper focuses on Internet of Things (IoT) architectures and knowledge generation out of streams of events as the primary elements concerning the creation of user-centric IoT services. We provide a general, symmetrical IoT architecture, which enables two-way bidirectional communication between things and users within an application domain. We focus on two main components of the architecture (i.e., Event Engine and Process Engine) that handle event transformation by implementing parametric Complex Event Processing (CEP). More specifically, we describe and implement the transformation cycle of events starting from raw IoT data to their processing and transformation of events for calculating information that we need in an IoT-enabled application context. The implementation includes a library of composite transformations grouping the gradual and sequential steps for transforming basic IoT events into business events, which include ingestion, event splitting, and calculation of measurements’ average value. The appropriateness and possibility of inclusion and integration of the implementation in an IoT environment are demonstrated by providing our implementation for a smart farming application domain with four scenarios that each reflect a user’s requirements. Further, we discuss the quality properties of each scenario. Ultimately, we propose an IoT architecture and, specifically, a parametric CEP model and implementation for future researchers and practitioners who aspire to build IoT applications.
An Event-based Microservice Platform for Autonomous Cyber-Physical Systems: The case of Smart Farming Dimitris Gkoulis, Cleopatra Bardaki, Elena Politi, Ioannis Routis, Mara Nikolaidou, George Dimitrakopoulos, Dimosthenis Anagnostopoulos 2021 16th International System of Systems Engineering Conference Sose 2021, 2021 Nowadays, we are realizing the Internet of Things capabilities as it has found many applications in the social and business life. Myriads of IoT devices have spread in everyday contexts supporting the collection and exchange of accurate, timely and complete information describing our products, cities, houses, farms, warehouses etc. The orchestration of the heterogeneous, connected IoT devices can support a variety of applications serving different business and everyday life purposes. This paper proposes a generic event-based platform that supports a variety of IoT-enabled application scenarios. To achieve scalability and reusability, the microservice paradigm is adopted. The proposed platform's architecture handles and composes microservices as atomic, autonomous and ephemeral functional entities that each reflects a specific application-oriented functionality. They are independent of the technology and the IoT infrastructure in the application domain. An IoT tier handles all the events and communicates the states of the IoT devices with the support of an Events communication channel. The functionality of the platform is demonstrated when supporting a smart farming case where the humidity of an IoT-enabled greenhouse is automatically controlled.
