Jonathan
Külz
Technische Universität München
Dienstort
Informatik 6 - Professur für Cyber Physical Systems (Prof. Althoff)
Boltzmannstr. 3(5607)/III
85748 Garching b. München
Curriculum Vitae
Jonathan Külz joined the Cyber Physical Systems Group as a doctoral student under the supervision of Prof. Dr.-Ing. Matthias Althoff in November 2021. He received his bachelor's degree in Mechatronics and Information Technology from the Karlsruher Institute of Technology (KIT) and his master's degree in Robotics, Cognition, Intelligence from the Technical University Munich (TUM).
His research includes:
- Algorithmic synthesis of modular robot compositions
- Model-based manipulator co-design
- Unifying benchmarks for modular and industrial robotics
- Computationally efficient methods for the assessment of robot capabilities
Are you looking for a thesis or a student project?
Disclaimer: Unfortunately, I will not be able to supervise new student projects until (including) May 2025.
I am always looking for self-motivated students that want to work in my research area. If you are interested in doing awesome research in (modular) robotics together, please follow these steps:
- Try to understand what my research is about (see below) and what exactly you are interested in.
- Have a look at this website to gain insight into my supervision approach and my expected student contributions.
- Write me a mail containing CV, transcript of records, and a concise statement of motivation.
- If I have capacities, we will most likely meet: Please prepare for the meeting by figuring out what you want to gain out of a thesis supervised by me.
Colloquially speaking, I care about the following questions:
- How do we design robots that are tailored to solving specific industrial tasks?
- How do we co-design robots and automation cells to perform well together?
- Which machine learning algorithms are well-suited to design the morphology of a robot?
- And last but not least: Considering that all of the above-posed questions require an extensive amount of simulation and evaluation: How can we assess "robot capabilities" computationally efficient for previously unseen robots?
Guided Research:
- Meta-Learning in Lexicographic Genetic Algorithms for Modular Robot Composition Optimization - Xintong He
Practical Courses:
- A hands-on introduction to Reinforcement Learning, SS 2024
Bachelor Theses
- Manipulation of a modular robot in construction - Mykhailo Razinkin
- Problem-Related Optimization of Discrete Robot Morphologies Using Reinforcement Learning - Zekun Jiao
- Hierarchical Filters for Modular Robot Morphology Fitness Evaluation - Denis Gretz
- Dynamic Parameter Identification for Modular Robots - Florian Mirkes
- Mastering the Game of Skat Using Decision Transformers - Sascha Benz
Master Theses
- Learning Robot Workspace Representations with Neural Fields - Xinyu Chen
- Robot-adaptive Neural Inverse Kinematics - Pramodkumar Choudhary
- Task-Based Modular Robot Configuration Synthesis with Deep Reinforcement Learning - Vinzenz Männig
Practical Courses
- Hybrid Reinforcement Learning for Modular Robot Optimization
- Model-Informed Reinforcement Learning for Optimizing Robot Design
- Supervised Learning for Robot Workspace Identification
Other
- Guided Research, Efficient Analytical Inverse Kinematics - Daniel Ostermeier
- Guided Research, Benchmarking Hybrid Reinforcement Learning for Robot Morphology Optimization - Pramodkumar Choudhary
- Internship, Dynamic Model Identification for Modular Robots - Justine Caulet
