Christian Osendorfer
Technical University of Munich, Leibniz University Hannover, Munich University of Applied Sciences
Papers
6
Total Citations
434
H-Index
5
About
Christian Osendorfer is a leading researcher in robot learning, reinforcement learning, and physical human-robot interaction. His foundational work on parameter-exploring policy gradients, notably the 2009 paper with 245 citations, introduced a powerful method for policy search that directly perturbs policy parameters rather than actions, enabling more efficient and stable control in continuous domains. This approach, further developed in his 2008 paper (60 citations), has become a cornerstone of modern reinforcement learning. Osendorfer has also made critical contributions to robot safety and tactile sensing. His 2015 study (68 citations) derived novel linear and non-linear features from physical contact models to discriminate between intended and unintended collisions, advancing safe human-robot interaction. In his 2014 work on model-free robot anomaly detection (42 citations), he developed algorithms to identify hardware or software failures without requiring explicit models, enhancing robot reliability. More recently, he explored computer vision for robotics, using convolutional neural networks and Gaussian processes to estimate finger grip forces from images (14 citations). Through these diverse contributions, Osendorfer has shaped both the theoretical foundations of policy gradients and the practical deployment of safe, perceptive robots.
Research Focus
Key Achievements
Top Papers
- 1Parameter-exploring policy gradients245 citations · 2009
- 2
- 3Policy Gradients with Parameter-Based Exploration for Control60 citations · 2008
- 4Model-free robot anomaly detection42 citations · 2014
- 5
- 6Training Neural Networks with Implicit Variance5 citations · 2013