European Agent Research
German, French, Swiss, UK, and EU-funded research on agentic AI systems.
European Agent Research
🇩🇪 German Research
DFKI (German Research Center for AI)
| Work | Year | Contribution |
|---|
| AJAN Framework | 2023 | SPARQL-extended Behavior Trees for Semantic Web agents, RDF knowledge modeling |
| Agentic AI for Robot Control | 2025 | LLM-based planning for Mobipick (indoor) + Valdemar (agricultural); demonstrates flexibility + fragility |
| Experience-Based Learning Systems | 2026 | Hybrid AI combining symbolic + subsymbolic; ProCAKE, CBRkit frameworks for agentic reasoning |
| Trusted AI + Agents | 2026 | “Third wave AI”: neural + symbolic; “Trust by Design” for certification |
Fraunhofer Institutes
| Work | Institution | Year | Contribution |
|---|
| Agentic AI Research | IAIS | 2025 | Reliability, transparency, domain fidelity for law/medicine/engineering |
| GenAI Multi-Agent Orchestration | FIT | 2024 | Morphological box for MAS orchestration; 4 propositions for organizational MAS |
| Multi-Agent Rule-Compliant Decisions | FOKUS | 2024 | Three-agent system (proposing + norm ID + fact retrieval) with Prolog reasoner |
Technical Universities
| Work | Institution | Year | Contribution |
|---|
| SAGE + OPACA | TU Berlin | 2025 | Dynamic tool discovery framework; zero-shot prompting; multi-agent extensibility |
| Memory-R1 | LMU/TU Munich/Darmstadt + Cambridge | 2025 | First RL framework for memory-augmented LLMs; 152 examples needed; ADD/UPDATE/DELETE/NOOP ops |
| MHAgentA | TU Clausthal | 2025 | Kahneman System 1/2 hybrid architecture; neurosymbolic combination |
| TB-CSPN | UniBW Munich | 2025 | Beyond prompt chaining architecture for agentic AI |
🇫🇷 French Research
INRIA
| Work | Year | Contribution |
|---|
| L’IA Agentique | 2025 | Multi-team overview: game theory, distributed systems, LLM orchestration, DB-agent links, human-robot interaction |
| SWE-Agents for Embodied Tasks | 2025 | SWE-agent adaptation for MiniGrid; source code vs interactive exploration comparison |
| Mimosa Framework (CNRS/Inria) | 2026 | Automatically synthesizes task-specific multi-agent workflows; iterative refinement via experimental feedback |
CNRS / Université de Lyon
| Work | Year | Contribution |
|---|
| SMA et agents autonomes génératifs | 2025 | Comprehensive French survey of LLM integration in MAS; open environment adaptation challenges |
ANR-Funded
| Project | Years | Contribution |
|---|
| LIAGORA (CentraleSupélec) | 2025-2029 | Trustworthy, frugal GenAI assistants; RAG optimization; multi-agent orchestration |
🇨🇭 Swiss Research
| Work | Institution | Year | Contribution |
|---|
| Society of HiveMind | ETH Zurich | 2025 | Multi-agent optimization imitating animal swarms; self-improvement via environment interaction |
| AgentsNet | ETH Zurich + RWTH Aachen + Google | 2025 | 100-agent coordination benchmark; distributed systems/graph theory inspiration |
🇬🇧 UK Research
University of Oxford
| Work | Year | Contribution |
|---|
| MALT | 2025 | Multi-agent LLM training: generation → verification → refinement with heterogeneous agents; value iteration credit assignment |
| Agora | 2024 | Meta-protocol for 100-agent networks; sidesteps “Communication Trilemma”; self-organizing networks |
| Task Complexity in MAS | 2025 | Framework characterizing tasks by depth (reasoning length) and width (capability diversity) |
| Rethinking MAS (Wooldridge) | 2025+ | Reframing multi-agent systems for LLM era; developing “Standard Model” for MAS |
University of Cambridge
| Work | Year | Contribution |
|---|
| Orchestrator | 2025 | Active inference for MAS; memory-driven collaboration; scalable reasoning with lightweight models |
| LLM-driven MAS for Digital Twins | 2025 | Leader + Reasoning + Member agents for infrastructure management |
🇮🇹 Italian Research
| Work | Institution | Year | Contribution |
|---|
| CARA | CNR Italy | 2025 | Consciousness-inspired agent architecture; BDI + Global Workspace Theory + Executive System |
🇪🇺 EU-Funded Projects
| Project | Budget | Years | Focus |
|---|
| AIXPERT | Horizon Europe | 2024-2028 | Explainable, accountable agentic AI; Agent-World Interface Layer + Cognitive Foundation Layer |
| HumAIne | Horizon Europe | 2022-2027 | Hybrid Human-AI decision support; Active/Swarm/Neuro-Symbolic Learning |
| MOSAICO | Horizon Europe | 2025-2029 | AI agent orchestration for software development |
| AIDA | €32M (EDF) | 2024-2028 | AI agents for cyber defence across land/air/sea/space |
| ELLIOT | Horizon Europe | 2024-2028 | Open, trustworthy multimodal foundation models |
| Next-Gen AI Agents | €38M | 2025 | Multi-agent frameworks for decentralized coordination |
| REGO | Horizon Europe | 2022-2026 | AI-powered micro-sized robot swarms |
| QuantAI | ERC Advanced | 2022-2027 | Classical/quantum learning agents for quantum error correction |
Key Research Strengths
| Country | Strength | Key Institutions |
|---|
| 🇩🇪 Germany | Hybrid AI, trust/safety, robotics | DFKI, Fraunhofer, TU Munich/Berlin |
| 🇫🇷 France | Multi-agent systems, LLM-MAS integration | INRIA, CNRS, Université de Lyon |
| 🇨🇭 Switzerland | Multi-agent optimization, benchmarks | ETH Zurich |
| 🇬🇧 UK | Agent protocols, communication, training | Oxford, Cambridge |
| 🇪🇺 EU | Large-scale funding, cross-institution, defense | Horizon Europe, EDF |