Manuel Kudruss
Papers
8
Total Citations
219
H-Index
5
About
Manuel Kudruss is a leading researcher in humanoid robotics and optimal control, whose work bridges the gap between theoretical optimization and real-world robotic motion. His primary research areas include nonlinear model predictive control (NMPC), whole-body motion generation, and learning-based control for complex robotic systems. Kudruss’s most impactful contribution is his pioneering demonstration that real-time NMPC can be implemented on position-controlled humanoid robots, as shown in his highly cited 2016 paper (97 citations), which introduced a reactive walking pattern generator that enables “walking without thinking.” He further advanced the field by developing a complete solution for fully-dynamic multi-contact motion generation (42 citations), expanding humanoid robots’ functional range beyond simple bipedal locomotion. Kudruss has also made significant strides in combining model-based and data-driven approaches, notably proposing a reinforcement learning framework to compensate for model-plant mismatch (37 citations) without requiring extensive real-world trials. His work on efficient derivative evaluation for rigid-body dynamics subject to kinematic constraints (2019) provides essential tools for gradient-based optimization in robotics. Through his research on the iCub and Leo humanoid platforms, Kudruss has demonstrated practical closed-loop control of walking and squatting motions, establishing himself as a key figure in making complex optimal control methods viable for real-time robotic applications.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3Model-Plant Mismatch Compensation Using Reinforcement Learning37 citations · 2018
- 4Benchmarking model-free and model-based optimal control28 citations · 2017
- 5
- 6
- 7
- 8