Johannes Silberbauer

Papers

Total Citations

H-Index

About

Johannes Silberbauer is a robotics researcher whose work sits at the intersection of classical mechanics and modern machine learning, specializing in differentiable physics and multi-body dynamics. His primary contribution is the development of a differentiable Newton-Euler algorithm, a framework that bridges the gap between rigid-body dynamics and gradient-based learning. This work, published in 2020, introduces a computation graph architecture that leverages the Lie Algebra form to efficiently encode the geometric structure of robot dynamics, enabling models that are both physically accurate and trainable via backpropagation. By making the Newton-Euler equations differentiable, Silberbauer enables robots to learn their own dynamics directly from data, moving beyond traditional model specification or simple linear regression. His follow-up work in 2021 extends this framework to real-world robotics applications, addressing practical challenges in deploying these hybrid models on physical hardware. With a combined 7 citations, his research is foundational for researchers working on model-based control, system identification, and simulation-to-real transfer, offering a principled way to embed physics into learning systems for more sample-efficient and robust robot control.