Program Schedule
Detailed agenda for the 3rd International Reinforcement Learning Bootcamp
Detailed agenda for the 3rd International Reinforcement Learning Bootcamp
A comprehensive look at our three-day event
Note: This program is tentative and subject to change as we finalize details.
Our program is designed to provide a comprehensive introduction to reinforcement learning, from basic concepts to advanced applications. The bootcamp spans three full days, with a mix of theoretical lectures, hands-on workshops, and networking opportunities.
Detailed agenda for September 16–18, 2026
Good to know: Participation in the bootcamp is free of charge. However, please note that travel, accommodation, lunches, and evening social events are at the participants' own expense.
Advance reservations are required for the social events. We will send out a separate email with detailed instructions on how to secure your spot.
| 11:00 – 12:00 | Arrival & Registration & Coffee |
| 12:00 – 13:30 | Opening Lunch (ARGE Beisl) (self-paid) |
| 13:30 – 14:00 | Opening and Introduction |
| 14:00 – 15:00 | Beginner Lecture: Introduction to Reinforcement Learning (No Math, No Panic) - Olga Mironova |
| 15:00 – 16:00 | Beginner Lecture: Reinforcement Learning Fundamentals - Simon Hirlaender |
| 16:00 – 16:30 | Coffee Break |
| 16:30 – 18:00 | Beginner Tutorial Hands-On Session: Tabular RL & Discrete MDPs |
| 18:30 – 21:00 | Welcome Dinner (ARGE Beisl) (self-paid) |
| 9:30 – 10:15 | Advanced Advanced Lecture: Policy Gradients and Actor Critics |
| 10:15 – 10:45 | Coffee Break |
| 10:45 – 11:30 | Keynote - Speaker TBA |
| 11:30 – 12:30 | Keynote - Speaker TBA |
| 12:30 – 14:00 | Lunch (ARGE Beisl) (self-paid) |
| 14:00 – 15:30 |
Intermediate
Tutorial Hands-On Session 2: Continuous Control & The Reality Gap
Environment: MuJoCo Ant → Crippled Ant
|
| 15:30 – 16:00 | Networking Break & Group Photo |
| 16:00 – 18:00 |
Advanced
Tutorial Hands-On Session 3: Custom MDP Design & Competition
Air Traffic Control Tournament
|
| 19:00 – 21:00 | Social Dinner (self-paid) |
| 9:30 – 11:00 | Advanced Interactive Wrap-Up Tutorial: Findings & Expert Panel |
| 11:00 – 11:30 | Coffee Break |
| 11:30 – 12:30 | Advanced Lecture: Agentic AI in the context of Reinforcement Learning |
| 12:30 – 13:00 | Final Round and Closing Remarks |
| 13:00 – 14:00 | Lunch (ARGE Beisl) (self-paid) |
| 15:00 – 17:00 | Salzburg City Visit |
Meet our expert speakers and instructors
University of Leoben, Austria
Associate Professor, Department of Mathematics and Information Technology
The overall goal of standard reinforcement learning is to find an optimal policy that maximizes the total reward. In this talk, I want to discuss the alternative objective of satisficing, which is content with performance above a given threshold. It will be argued that this more modest aim is not only often more suitable in applications but also can be shown to fulfill stronger performance guarantees.
Ronald Ortner is an associate professor at the Department of Mathematics and Information Technology at the University of Leoben, Austria, where he has served since 2003 (assistant professor) and since 2010 as associate professor, interrupted by a sabbatical at Inria Lille, France in 2012. His research focuses on questions in theoretical reinforcement learning, particularly regret analysis in settings ranging from multi-armed bandits to general Markov decision processes.
University of Alberta, Canada
Fellow & Canada CIFAR AI Chair at Amii, Full Professor, Research Scientist at Google DeepMind
The talk title and abstract will be announced soon. Stay tuned!
Michael Bowling is a Fellow and Canada CIFAR AI Chair at Amii, a full professor at the University of Alberta, and a Research Scientist at Google DeepMind. He is best known for his groundbreaking work in poker AI, including Cepheus (2015) and DeepStack (2016) - both published in Science - which achieved landmark results in imperfect-information games. He also led the development of the Arcade Learning Environment, which was instrumental in establishing deep reinforcement learning as a research field. He is a principal investigator in the RLAI Lab and leader of the Computer Poker Research Group at the University of Alberta.
European Organization for Nuclear Research (CERN), Switzerland (remote)
Accelerator Physicist, Team Leader - Super Proton Synchrotron & Low Energy Ion Ring
The talk title and abstract will be announced soon. Stay tuned!
Verena Kain is an accelerator physicist at CERN, supervising students and postdocs and leading the teams running the Super Proton Synchrotron and Low Energy Ion Ring. She brings extensive experience in commissioning and operating large-scale accelerator facilities, having served as one of the Engineers in Charge during the commissioning and operation of the Large Hadron Collider (LHC) from 2007 to 2013. Her responsibilities include software development for the control room - combining physics and hardware - and she is actively pushing for automation to achieve higher efficiency, reproducibility, and increased flexibility of CERN's high-energy frontier machines.
University of Leoben, Austria
Chair of the Cyber-Physical-Systems Institute, 2019 German Young Researcher Award
Humanoid robots are expected to play an increasingly important role in industrial and consumer environments. While recent advances have demonstrated promising results for pick-and-place tasks, achieving robust, data-efficient, and transferable robot skills remains a significant challenge. This talk presents learning-based approaches for humanoid robot manipulation that combine data augmentation, motion capture, and whole-body teleoperation using AR/VR systems and wearable motion trackers. The resulting policies are trained end-to-end using reinforcement learning frameworks, including diffusion policies and flow-matching approaches, targeting demanding industrial scenarios such as recycling operations, handling toxic powders, and manipulating hazardous batteries.
Elmar Rueckert is a full professor and chair of the Cyber-Physical-Systems Institute at the University of Leoben, Austria. He received his PhD in computer science from Graz University of Technology in 2014 and previously held positions as senior researcher and research group leader at TU Darmstadt and as assistant professor at the University of Lübeck. His research spans stochastic machine and deep learning, robotics, reinforcement learning, and human motor control. In 2019, he was awarded the German Young Researcher Award.
Intensive learning experiences combining theory and practice
Learn the foundational theories of Reinforcement Learning. Formulate discrete Markov Decision Processes (MDPs) and solve them using classic tabular methods in an interactive Google Colab notebook environment.
Environment: Maze (Colab)
Take the leap into continuous action spaces. Explore the complexities of continuous control policies in simulation and study the challenges of transferring policies from MuJoCo Ant to modified physics engines (Crippled Ant).
Environment: MuJoCo Ant → Crippled Ant
Design custom reward functions, observation spaces, and transition dynamics. Apply advanced techniques to optimize agents for air traffic control, and compete against other participants in the simulation tournament.
Environment: UMFlightEnv (air-traffic)
The intermediate and advanced sessions share one codebase, the AirTraffic package, documented in full in the AirTraffic usage guide.
