Simon Hirländer
Vice Director, IDA Lab | Team Lead, Smart Analytics & Reinforcement Learning
Paris Lodron University Salzburg | Visiting Scientist, CERN
About
I am Vice Director of the IDA Lab and lead the Smart Analytics & Reinforcement Learning (SARL) team at PLUS University in Salzburg.
Operating as a Theoretical Physicist and Decision Scientist, I bridge the gap between rigorous theoretical modeling and modern Artificial Intelligence. My approach integrates Hamiltonian mechanics, potential theory, and control theory with advanced machine learning to solve complex dynamical problems.
Pioneering AI in Control: In 2018, I led the first Deep Reinforcement Learning experiment at the CERN accelerator complex, establishing the foundational baseline for autonomous particle accelerator control. Today, I continue this mission as a Visiting Scientist at CERN, developing safe, uncertainty-aware RL agents for the world's most critical machines.
Beyond foundational research, I aggressively pursue technology transfer, applying high-precision control methods to industrial challenges in steel interaction, medical robotics, and smart energy grids.
Research Vision: To replace "black box" approximations with rigorous, physics-aware AI that guarantees safety and optimality in complex physical systems.
Research Highlights
Recent News
IPAC 2026 Contributions
Presenting 4 posters at IPAC 2026, including RL beyond greedy optimization and Causal GP-MPC.
Read more →INSPIRE Project Starts
Launch of the INSPIRE project: AI prediction and personalized instruction for STEM education.
Read more →Background
The Smart Analytics & Reinforcement Learning (SARL) research group operates as an interdisciplinary powerhouse, fusing concepts from Theoretical Physics, Control Theory, and Advanced Artificial Intelligence. Our research is not limited to a single domain but focuses on the underlying mathematical structures that govern complex systems—whether they are particle beams, industrial assembly lines, or autonomous robots.
Methodological Expertise: Our research stack spans the full spectrum of modern decision science. We specialize in Model-Based & Hierarchical Reinforcement Learning for high-dimensional control, Gaussian Process MPC for safe, uncertainty-aware operations, and Causal Inference to disentangle structure from noise. Our toolkit extends to Unsupervised Anomaly Detection, Fractal Time-Series Analysis, and Neuro-Symbolic AI, allowing us to tailor the mathematical architecture to the specific physics of the problem—whether for the LHC, a smart factory, or autonomous maritime vessels.
PhD Thesis: Exact Solutions for Accelerator Physics
Solving the "Factor of Two" Discrepancy at the CERN LHC via the Riemann-Sphere
The Challenge: For years, the Large Hadron Collider (LHC) operated with a theoretical blind spot: predictions for the beam's stability consistently deviated from measurements by a factor of two. Standard numerical approximations failed to capture the complex electromagnetic reality of the beam screen.
The Novel Framework: Resolving this discrepancy required a fundamental paradigm shift. I introduced a novel mathematical framework by reformulating the problem on the Riemann-sphere. This approach represents a synthesis of highly abstract potential theory and practical machine engineering, demonstrating high mathematical originality to solve otherwise intractable physical problems with exact precision.
Key Contributions
- Conceptual Breakthrough: Pioneered the "Riemann-Sphere Framework" for accelerator physics, handling both open and closed boundary conditions.
- Mathematical Discovery: Derivation of a novel integral representation of the Neumann function on the Riemann-sphere, enabling the first exact solutions to complex magnetostatic boundary problems (including unbounded star-like domains).
- The "Rect-Ellipse" Solution: A closed-form analytical model using Jacobian elliptic functions that exactly maps the LHC beam screen.
- Rigorous Safety Proofs: Proved the "Polygonal Error Bound Theorem," providing engineers with mathematically guaranteed safety margins.
The Impact: This rigorous approach resolved the LHC tune-shift discrepancy, matching measurements with >99% accuracy. The work now underpins stability predictions for the High-Luminosity LHC (HL-LHC).
Community Leadership
- Organize the "RL Coffee" - monthly meetup (first Friday of each month) for RL researchers and practitioners
- Co-organize the RL Bootcamp - annual intensive training program
- Lead the RL4AA (Reinforcement Learning for Autonomous Accelerators) international collaboration
- Member of the Supervision Team for the Doctorate School PLUS (DSP) "Human-Centered AI", advancing interdisciplinary AI research
- Delivered a TEDx talk: "AI is what we make it"
Collaborations
Research Institutions: CERN, DESY, KIT, SLAC, GSI, DLR, University of Oxford, University of Malta, Stanford University, LANL, AI Austria
Industry Partners:
- Medical Technology: W&H Dentalwerk (Precision control for piezo drivers)
- Steel & Heavy Industry: Danieli Automation (Rolling mill optimization), Palfinger (Hydraulics)
- Energy & Sustainability: Copa-Data (Smart grid optimization)
- Manufacturing: STIWA Group (Assembly line automation)
- Knowledge Systems: Redlink (Neuro-symbolic AI)
- FinTech: JUIX (Automated auditing)
Teaching
Courses at Paris Lodron University Salzburg (PLUS):
- Advanced Reinforcement Learning and Agentic AI Systems
- Introduction into Deep Reinforcement Learning
- Introduction to Reinforcement Learning
- Advanced Topics in Reinforcement Learning
- Special Topics in Data Science
- Mathematical Foundations in Precision Medicine
- AI Bachelor Seminar
Current Students
I am privileged to work with talented students on exciting research topics. Our team includes PhD students, Master students, and Bachelor students working on various aspects of RL and AI.
Recent topics include: Multi-agent RL, energy optimization, accelerator control, nautical robotics, assembly line optimization, and more.
Contact
Email: simon.hirlaender@plus.ac.at
Office: IDA
Lab, PLUS University Salzburg
Department: Artificial Intelligence and Human Interfaces
Faculty: Digital
and Analytical Sciences
Interested in collaboration? Feel free to reach out if you're interested in research collaborations, industrial partnerships, or student opportunities.