Industrial Domains

Space

The Space Industrial domain focuses on creating applications where AI can leverage remote sensing data to develop systems in the space sector that are easily understood by regulatory bodies and adaptable to constantly changing scenarios. There has been growing interest in space development, providing researchers with access to space-based Earth Observation (EO) capabilities to capture critical environmental data, improving accuracy in areas like agriculture and natural disaster management. 

By combining EO with AI, the space industry enables large-scale, near-real-time monitoring of Earth’s surface activities, helping predict and track events such as natural disasters, disease outbreaks, and illegal activities.  

Additionally, it helps identify key correlations between environmental processes and satellite-derived data, offering valuable insights for decision-making and sustainable resource management. 

Sub-Domains
Use Cases
Projects
Project: SeaScope: Explainable AI for EO-Driven Maritime Analysis & Monitoring

Third parties involved: Ctrl and NKUA

Maritime EO analysis is still underexploited despite growing satellite data. Researchers rely on custom code for each sensor and use case. Rule-based tools don’t scale, lack transparency and delay insights into marine events. SeaScope tackles these issues with an explainable AI agent to let researchers analyse events via natural language, e.g. “Map oil spills and vessel movement in May 2025” and get results on an interactive map. Using vision LLMs, Retrieval-Augmented Generation and Google Earth Engine, SeaScope auto-generates, runs, and explains EO workflows. Within six months, SeaScope will reach TRL 6 by implementing a pilot on vessel detection related to oil spills, water, and air quality, with a focus on the Cyprus EEZ and Mediterranean Sea.   

Project: SeaScope: NeuVNav: Neuro-Evolving Visual Navigation for Low SWaP UAVs

Third parties involved: H-AERO

NeuVNav addresses the critical challenge of robust, autonomous, vision-based navigation for low SWaP UAVs. It proposes a novel, bio-inspired hybrid AI architecture that combines a lightweight Convolutional Neural Network (CNN) for efficient feature extraction with an evolving Spiking Neural Network (SNN) for ultralow-power, event-driven path planning and obstacle avoidance. The system will feature onboard neuroevolution, enabling it to adapt its navigation logic in real-time to dynamic environments. This project will deliver trusted, computationally frugal AI models, a proof-of-concept demonstrator on our h-aero platform, and a high-impact publication, directly advancing ENFIELD’s Green and Trustworthy AI pillars.   

Project: Quantum-Inspired Tensor Networks for On-Board Hyperspectral Data Processing

Third parties involved: Multiverse Computing SL 

The project aims to develop a lite AI model capable of running in edge processors, following a multi-model approach, to tackle the challenge of on-board hyperspectral data processing for cloud and cloud shadow masking. Vision Transformers (ViTs), Spectral-Spatial Graph Neural Networks (GNNs), and hybrid CNN-Transformer models will be integrated to enhance feature extraction and classification. Activities include analysing feature selection techniques, model development, validation, and optimisation. Expected outcomes are improved accuracy in cloud and cloud shadow detection and suitability for on-board deployment, contributing to ARD hyperspectral data delivery for European hyperspectral satellite missions.   

Project: HYPERBOLA: HYPerspEctRal onBOard cLoud AI

Third parties involved: Neuralio OÜ 

The project aims to develop a lite AI model capable of running in edge processors, following a multi-model approach, to tackle the challenge of on-board hyperspectral data processing for cloud and cloud shadow masking. Vision Transformers (ViTs), Spectral-Spatial Graph Neural Networks (GNNs), and hybrid CNN-Transformer models will be integrated to enhance feature extraction and classification. Activities include analysing feature selection techniques, model development, validation, and optimisation. Expected outcomes are improved accuracy in cloud and cloud shadow detection and suitability for on-board deployment, contributing to ARD hyperspectral data delivery for European hyperspectral satellite missions.