Recipient organisationUniversity of ReadingSource-published name: University of Reading
Funding£268K
PeriodOct 2025 — Oct 2029
In plain English
AI plain-English summary
A doctoral network will train 15 researchers to build artificial intelligence that runs directly on devices like robots, sensors, and medical monitors, rather than in the cloud. Embedded AI currently faces fundamental limits: models are too large and power-hungry for small devices, struggle to learn across distributed networks, and lack the transparency needed for safety-critical use. This project addresses those gaps by designing low-footprint AI models that operate under tight energy and bandwidth constraints, developing distributed learning methods for heterogeneous device networks, and building explainability and security into the systems from the ground up. If successful, the research could make autonomous robots, underwater Internet-of-Things sensors, mobile health monitors, and smart farming equipment more capable and trustworthy without requiring constant internet connections. The trained researchers will move into industry, academia, or government, strengthening Europe’s talent base in a technology that quietly underpins everything from factory automation to environmental monitoring. The project is applied and industry-focused, with a clear pathway from fundamental methods to real-world deployment.
View original technical description
Embedded Artificial Intelligence (AI) has emerged as a transformative technology with immense potential to revolutionise various domains, spanning from robotics and healthcare to environmental monitoring and the Internet of Things. This Doctoral Network (DN) project ANT aims to train a network of 15 excellent Doctoral Candidates (DCs) by addressing the fundamental challenges of Embedded AI and accelerating the development of Embedded AI systems and applications through an innovative and interdisciplinary research and training program. ANT consists of four interconnected Work Packages (WPs) that encompass different aspects of Embedded AI. WP1 tackles the challenges in designing low-footprint standalone Embedded AI models under resource constraints and with diverse contexts and evolving environments. WP2 goes beyond standalone Embedded AI and designs innovative distributed and scalable learning solutions for heterogeneous Embedded AI networks under energy and bandwidth constraints. WP3 enhances the trustworthiness of Embedded AI with explainability, robustness, security, and privacy. ANT concludes in WP4 with a concerted effort to transfer fundamental research contributions to industry-relevant applications in autonomous robotics, underwater IoT, mobile healthcare, and smart farming, boosting Europe's position in the global AI market both from a talent and a technological perspective. These interdisciplinary and inter-domain research training, along with the comprehensive soft-skills training (spanning from presentation skills to intellectual property, marketing, and economics, etc.) will make ANT's 15 DCs highly employable in various industries, academia, or public government bodies, and will position the EU at the forefront of the emerging revolution of Embedded AI on Things.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know