Periodic Reporting for period 1 - REMARKABLE (Rural Environmental Monitoring via ultra wide-ARea networKs And distriButed federated Learning)
Okres sprawozdawczy: 2023-01-01 do 2024-12-31
The REMARKABLE project addresses a critical global challenge: reducing the urban-rural digital divide by enabling transformative digital transformation in rural areas through Internet of Things (IoT) and Machine Learning (ML) technologies. Rural regions, which cover significant portions of Europe and Africa, suffer from limited connectivity, stalling technological advancement and economic opportunities. By leveraging IoT and ML tools, the project aims to extract actionable insights from environmental data to improve societal and economic outcomes in these regions. The African continent is highlighted for its potential impact due to its vast uncultivated agricultural land and underdeveloped digital infrastructure.
At the core of REMARKABLE’s vision is the development of energy-efficient and secure IoT/ML systems tailored for deep rural applications. These systems will support diverse use cases, such as wildlife conservation, precision agriculture, and pollution monitoring. The project is built upon an interdisciplinary collaboration between academic and industrial partners from Europe and Africa, ensuring holistic and innovative solutions. Demonstration sites across both continents will serve as real-world testbeds for deploying and validating these technologies.
The project’s pathway to impact focuses on fostering sustainable research and innovation ecosystems. By integrating state-of-the-art IoT sensors, ultra-wide area networks, and distributed data analytics models, REMARKABLE aims to bring tangible benefits to rural communities. It will enable farmers to improve yields through precision agriculture, safeguard endangered wildlife, and enhance rural tourism experiences. Furthermore, by developing partnerships with industrial stakeholders, the project ensures its outputs have direct routes to commercialization, driving both technological advancements and economic growth.
Finally, REMARKABLE prioritizes capacity building and knowledge exchange through interdisciplinary training, joint research, and public outreach. By involving local stakeholders and leveraging the unique geographical and socio-economic characteristics of African and European regions, the project ensures its sustainability and relevance. These initiatives will strengthen long-term research collaborations, empower a new generation of researchers with interdisciplinary expertise, and foster innovative solutions that address pressing global challenges.
At the core of REMARKABLE’s vision is the development of energy-efficient and secure IoT/ML systems tailored for deep rural applications. These systems will support diverse use cases, such as wildlife conservation, precision agriculture, and pollution monitoring. The project is built upon an interdisciplinary collaboration between academic and industrial partners from Europe and Africa, ensuring holistic and innovative solutions. Demonstration sites across both continents will serve as real-world testbeds for deploying and validating these technologies.
The project’s pathway to impact focuses on fostering sustainable research and innovation ecosystems. By integrating state-of-the-art IoT sensors, ultra-wide area networks, and distributed data analytics models, REMARKABLE aims to bring tangible benefits to rural communities. It will enable farmers to improve yields through precision agriculture, safeguard endangered wildlife, and enhance rural tourism experiences. Furthermore, by developing partnerships with industrial stakeholders, the project ensures its outputs have direct routes to commercialization, driving both technological advancements and economic growth.
Finally, REMARKABLE prioritizes capacity building and knowledge exchange through interdisciplinary training, joint research, and public outreach. By involving local stakeholders and leveraging the unique geographical and socio-economic characteristics of African and European regions, the project ensures its sustainability and relevance. These initiatives will strengthen long-term research collaborations, empower a new generation of researchers with interdisciplinary expertise, and foster innovative solutions that address pressing global challenges.
During the reporting period, the REMARKABLE project made significant technical progress. A secure and scalable IoT platform for rural applications was developed, integrating lightweight ML models and decentralized data-driven methods. TinyML-based solutions for bird identification and animal tracking were implemented, along with UAV-assisted secure localization for GPS-denied environments. Initial digital twin modeling integrated real-world sensor and camera data.
In ultra-wide area IoT networks, the project investigated LP-WAN limitations and designed UAV-assisted network extensions, demonstrating successful prototypes. Research into deep rural connectivity via high-altitude platforms and LEO satellites progressed, laying the foundation for future trials. The evolving network architecture prioritizes energy efficiency, communication reliability, and multi-connectivity integration.
The distributed data analytics component introduced novel methods for efficient IoT data processing. Feature extraction and dimensionality reduction improved real-time sensor data processing. Federated learning architectures were developed to enable privacy-preserving model training across distributed IoT nodes. Intelligent observatories for species identification and behavior recognition were deployed, demonstrating practical applications.
Validation activities spanned multiple pilot sites. In Spain, IoT sensors and digital twin frameworks were deployed in wetlands for ecosystem management. Serbia tested smart buoys, UAV-based IoT connectivity, and air quality monitoring, validating LoRaWAN and NB-IoT effectiveness. Nigeria implemented IoT-based agricultural solutions, including soil monitoring, drone-assisted farming, and aquaculture testbeds. Morocco developed a mobile application for almond tree disease detection. South Africa deployed a large-scale Eucalyptus monitoring system, integrating low-cost IoT sensors with cloud-based analytics.
Key outcomes include 12 deliverables and successful pilot demonstrations, validating the technologies' impact on environmental monitoring and precision agriculture. The project remains on track, focusing on refining ML models, optimizing network performance, and extending real-world validations.
In ultra-wide area IoT networks, the project investigated LP-WAN limitations and designed UAV-assisted network extensions, demonstrating successful prototypes. Research into deep rural connectivity via high-altitude platforms and LEO satellites progressed, laying the foundation for future trials. The evolving network architecture prioritizes energy efficiency, communication reliability, and multi-connectivity integration.
The distributed data analytics component introduced novel methods for efficient IoT data processing. Feature extraction and dimensionality reduction improved real-time sensor data processing. Federated learning architectures were developed to enable privacy-preserving model training across distributed IoT nodes. Intelligent observatories for species identification and behavior recognition were deployed, demonstrating practical applications.
Validation activities spanned multiple pilot sites. In Spain, IoT sensors and digital twin frameworks were deployed in wetlands for ecosystem management. Serbia tested smart buoys, UAV-based IoT connectivity, and air quality monitoring, validating LoRaWAN and NB-IoT effectiveness. Nigeria implemented IoT-based agricultural solutions, including soil monitoring, drone-assisted farming, and aquaculture testbeds. Morocco developed a mobile application for almond tree disease detection. South Africa deployed a large-scale Eucalyptus monitoring system, integrating low-cost IoT sensors with cloud-based analytics.
Key outcomes include 12 deliverables and successful pilot demonstrations, validating the technologies' impact on environmental monitoring and precision agriculture. The project remains on track, focusing on refining ML models, optimizing network performance, and extending real-world validations.
Key achievements in the IoT platforms for rural areas include the deployment of tinyML algorithms for edge-based animal monitoring, addressing computational constraints and intermittent connectivity. The project advances real-time behavior recognition in poultry farming, enhancing farm management and sustainability. Secure localization methods were developed to mitigate node malfunctions, incorporating min-max and GTRS frameworks alongside ADMM-based decomposition techniques. A multi-modal data fusion approach integrates sensor measurements and wireless signals, improving localization reliability in constrained environments. Digital Twin system modeling utilizes collected sensor and camera data for ML training and environmental simulations. Current efforts focus on integrating eucalyptus tree measurements and bird observation data to enhance real-time monitoring and scalability for rural IoT applications.
In the distributed data analytics area, emphasis is placed on designing features for data-efficient machine learning, secure and scalable federated learning, and minimizing communication and computation requirements. Novel methods beyond state-of-the-art have been developed to extract relevant features from the unstructured data. Also, feature dimensionality reduction methods are being developed to minimize the memory space required to store the data for machine learning model training in edge or fog nodes.
Several advances with potential impact have been developed in all the pilot sites, including sensor networks, intelligent observatories, localization algorithms, water and air quality monitoring systems, Cellular IoT coverage extension using UAVs. Aeroponic systems. Enhanced LoraWAN based acquisition system to cite some of them. However, all these developments and advances need to be tested for a large period for a rigorous validation process.
In the distributed data analytics area, emphasis is placed on designing features for data-efficient machine learning, secure and scalable federated learning, and minimizing communication and computation requirements. Novel methods beyond state-of-the-art have been developed to extract relevant features from the unstructured data. Also, feature dimensionality reduction methods are being developed to minimize the memory space required to store the data for machine learning model training in edge or fog nodes.
Several advances with potential impact have been developed in all the pilot sites, including sensor networks, intelligent observatories, localization algorithms, water and air quality monitoring systems, Cellular IoT coverage extension using UAVs. Aeroponic systems. Enhanced LoraWAN based acquisition system to cite some of them. However, all these developments and advances need to be tested for a large period for a rigorous validation process.