Michael Grner
Papers
1
Total Citations
16
H-Index
1
About
Michael Grüner is a roboticist whose work focuses on bridging the gap between simulation and real-world manipulation, particularly in the domain of planar pushing. His key research areas include data-driven dynamics modeling, reinforcement learning, and sim-to-real transfer for robotic manipulation. Grüner’s major contribution lies in developing self-adapting recurrent models that learn object-pushing dynamics entirely in simulation, yet generalize robustly to physical hardware without retraining. This approach addresses the fundamental instability of analytical physics models by using learned representations that adapt to unmeasured parameters like friction and inertia. His most cited work, "Self-Adapting Recurrent Models for Object Pushing from Learning in Simulation" (2020), has accumulated 16 citations and demonstrates a novel method for achieving zero-shot transfer from simulation to reality. By enabling robots to push objects accurately without requiring precise physical parameters, Grüner’s research advances practical applications in warehouse automation, assistive robotics, and industrial manipulation. His work exemplifies how deep learning can overcome the limitations of traditional physics-based control, offering a scalable path toward more dexterous and adaptive robotic systems.
Research Focus
Key Achievements
Top Papers
- 1