Papers
8
Total Citations
477
H-Index
8
About
Felix Berkenkamp is a prominent researcher at the intersection of safe machine learning, reinforcement learning, and robotics, whose work addresses one of the field's most pressing challenges: enabling autonomous systems to learn efficiently while guaranteeing safety. His research has made foundational contributions to safe exploration in reinforcement learning, demonstrating how Gaussian processes and Bayesian optimization can be harnessed to prevent unsafe actions during learning — a critical requirement for real-world robotic applications. His 2016 paper on safe exploration in finite Markov decision processes (68 citations) laid important theoretical groundwork, while his work on Bayesian optimization with safety constraints (80 citations) translated these ideas into practical parameter tuning frameworks for robotics. Berkenkamp has also advanced stability certification for learned controllers, notably through the Lyapunov Neural Network approach (69 citations), which provides formal safety guarantees for adaptive systems. His innovative strategy of intelligently trading off simulations and physical experiments (111 citations) has proven highly influential in making reinforcement learning more sample-efficient and deployable. With over 470 cumulative citations across his key works, Berkenkamp's research offers indispensable tools for researchers seeking to bridge the gap between theoretical safety guarantees and practical autonomous systems.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4Safe Exploration in Finite Markov Decision Processes with Gaussian Processes68 citations · 2016
- 5
- 6
- 7
- 8