Papers
9
Total Citations
253
H-Index
5
About
Maximilian Karl is a robotics and machine learning researcher whose work spans reinforcement learning, robot control, and generative modeling, with a particular focus on enabling robots to learn complex skills efficiently from high-dimensional sensory data. His most influential contribution, "Stable Reinforcement Learning with Autoencoders for Tactile and Visual Data" (2016, 142 citations), demonstrated how autoencoders could make reinforcement learning tractable for robots operating with rich tactile and visual feedback — a foundational step toward more autonomous robotic systems. Complementing this, his work on Dynamic Movement Primitives in latent spaces of variational autoencoders (57 citations) advanced how robots can generalize learned movements across complex, high-dimensional settings. Karl has further contributed to model-based reinforcement learning for drone control, systematic analyses of action spaces in sim-to-real transfer, and intrinsic motivation through empowerment. His earlier research on compliant, worm-like continuum robots reflects a grounding in safe human-robot interaction. More recently, his work on language-informed multi-task visual world models signals a forward-looking interest in scalable, generalizable robot learning. Across more than 250 cumulative citations, Karl's research consistently addresses the challenge of making reinforcement learning both sample-efficient and practically deployable in real-world robotic applications.
Research Focus
Key Achievements
Top Papers
- 1Stable reinforcement learning with autoencoders for tactile and visual data142 citations · 2016
- 2
- 3Learning to Fly via Deep Model-Based Reinforcement Learning24 citations · 2020
- 4
- 5
- 6
- 7Design of an inherently safe worm-like robot2 citations · 2013
- 8Efficient Empowerment2 citations · 2015
- 9LIMT: Language-Informed Multi-Task Visual World Models1 citations · 2025