Artificial Intelligence (AI) technologies are crucial for ongoing digital transformation across various business sectors and critical infrastructures. The rise of the Internet of Things (IoT) and the IoT-Edge-Cloud continuum has led to the generation of vast amounts of sensor data. This data must be efficiently processed and maintained to address strategic challenges in different industrial domains, serving as a key resource for AI and a critical driver of global economic development.
However, AI-driven IoT (AIoT) systems and their associated services and applications require large amounts of real data at scale to enhance their accuracy, robustness, and sustainability. Despite the importance of real data, its availability is often limited due to several constraints. Even when real data is available, it is often not in the required format or lacks the necessary labeling and annotations for AIoT systems. Furthermore, the data is frequently imbalanced, with nominal system behavior overrepresented and data on rare or anomalous events scarce. AIoT systems must also be able to adapt to the dynamic nature of smart spaces, particularly in critical infrastructure ecosystems, requiring ongoing adjustments to account for un-modeled phenomena and maintain robustness during operation.
In this context, the mission of the PANDORA project (https://pandora-heu.eu/) is to develop a comprehensive AI-based and domain-informed framework, which aims to optimize the preparation and the delivery of complete and trustworthy datasets for training and enhancing AI models deployed in AIoT systems (Phase 1). Furthermore, PANDORA aims to increase the degree of the autonomy, trustworthiness and energy efficiency of the relevant processes for designing these AI models and managing and using the respective IoT-enabled datasets in smart space ecosystems (Phase 2). The overall PANDORA vision is presented in the figure below (source: https://pandora-heu.eu/):
PANDORA unites key innovators and technology experts from the AI, data, and robotics ecosystems to develop advanced solutions for AIoT systems, including 25 project partners overall. Among them, University of Novi Sad, Faculty of Sciences (UNSPMF) contributes as a project partner, while the project is coordinated by the National Technical University of Athens (NTUA). Our multidisciplinary team focuses on creating energy-efficient data sensing mechanisms from IoT infrastructures, human-centered data reasoning, and customizable, trustworthy datasets for model training and testing before deploying AIoT systems. Additionally, PANDORA emphasizes robust learning techniques to ensure autonomous and continuous AIoT operations by efficiently deploying data pipelines for knowledge inference and model optimization.
The PANDORA framework will be tested in real-life scenarios to tackle industrial challenges, specifically focusing on verifiable and validated ML-driven data pipelines for AIoT systems. These systems will support the digital strategies and operational needs of manufacturing spaces, conventional buildings, and critical energy infrastructures. Five trial scenarios will demonstrate the framework’s ability to optimize AI model design and provide end-to-end support for “data for AI” pipelines across cloud computing environments.
Through these efforts, PANDORA aims to contribute to research, industrial, and standardization initiatives. The project highlights the importance of AI-based, ICT-enabled approaches in optimizing data preparation and usage processes. By advancing trustworthiness and efficiency, PANDORA sets new standards for continuous and adaptive operations within AIoT ecosystems, making these principles the foundation for developing “data for AI” pipelines.
Vladimir Kurbalija and Dusan Jakovetic, Department of Mathematics and Informatics, Faculty of Sciences, Novi Sad