Cognitive Systems & Robotics Laboratory

Generative Robotics and AI for EU Industrial Leadership

Generative Robotics and AI for EU Industrial Leadership

GRAIL is a European project developing the next generation of trustworthy, human-centric, and sovereign Generative AI (GenAI) solutions for robotics. The project addresses key barriers to the large-scale adoption of AI-powered robotics in manufacturing, including the limited availability of high-quality robot data and the need for compliance with European regulations and safety requirements. GRAIL introduces an integrated agentic architecture that combines specialised AI agents for planning, grounding, and control through a shared knowledge representation. This approach enables continuous learning, efficient knowledge reuse, and skill transfer across diverse robotic platforms. Human oversight is embedded through Human-in-the-Loop (HITL) and Human-on-the-Loop (HOTL) mechanisms, allowing operators to understand, monitor, and intervene in robot decision-making processes. Predictive world models, deterministic fallback strategies, and runtime safeguards ensure safe, adaptive, and reliable operation. The project will demonstrate its technologies across five industrial sectors (automotive, aerospace, intralogistics, steel, and electronics) while leveraging federated learning and high-performance computing to promote scalability, sustainability, and European technological sovereignty.

CERTH/ITI contributes to the technical implementation of GRAIL by developing core technologies for trustworthy and human-centric Generative AI in robotics. Its research activities focus on multimodal grounding, active perception, uncertainty estimation, expert routing mechanisms, and multi-scale memory architectures that enable robust perception across diverse industrial use cases. In parallel, it leads the development of an ethical safeguard layer that monitors AI-generated plans and actions at runtime, ensuring compliance with ethical principles and supporting trustworthy, explainable, and safe human–robot collaboration. CERTH/ITI also contributes to the co-design, integration, and iterative evaluation of the GRAIL framework by defining validation methodologies and performance indicators that assess system adaptability, usability, safety, trust, and resource efficiency under realistic industrial conditions. As technical coach for the intralogistics use case, CERTH/ITI supports requirements elicitation, system integration, and collaboration between technology providers, end users, and third-party innovators, facilitating the alignment of research outcomes with industrial needs.