Bachelor Thesis in Mathematics, Physics or Computer Science “Dynamic Filter for Walking Motion Correction“
The humanoid robot walking and avoiding obstacles.
Comparison of reference, real and corrected center of pressure trajectories.
Abstract
The ability of autonomous bipedal walking for humanoid robots is still an open research topic for the robotics community. The current state of the art uses extensive model simplifications of the robot's dynamics in the form of inverted-‐pendulum-‐like models. The planned motion using a simplified model can lead to problems during its execution on the robot due to neglected effects of the dynamics, most often the change of angular momentum. A common way to "smooth" out these trajectories is achieved by a so-‐called dynamic filter that utilizes a given whole-‐body dynamic model of the robot. This leads to much smoother and more reliable motions on the robot, cf. [1].
Project The scope of the project is suitable for a beginner software practical or BSc thesis. The project covers: • Implementation of a dynamic filter as proposed by „Kajita“ in [2] • Comparison of dynamic filter to Newton-‐Raphson iteration • Short paper/BSc thesis summarizing approaches and results
Requirements
• • •
Good knowledge of Analysis I & II, Linear Algebra I Knowledge of Numerical Optimization is helpful Programming experience with Python is helpful
References
[1] „NMPC-‐based continuous reactive walking pattern generator“, Naveau et al., IEEE Robotics and Automation Letters, 2016. [2] „Biped walking pattern generation by using preview control of zero-‐moment point“, Kajita et al., in the proceedings of IEEE International Conference on Robotics and Automation, 2003.
Contact
Prof. Dr. Katja Mombaur,
[email protected]‐heidelberg.de Manuel Kudruss,
[email protected]‐heidelberg.de www.orb.uni-‐hd.de
