PhD Defense: Bio IK: A Memetic Evolutionary Algorithm for Generic Multi-Objective Inverse Kinematics

20 August 2020

Inverse kinematics is an essential task in robotics for solving motion control, manipulation as well as different interaction scenarios. In this talk, I will present an efficient, scalable and customizable memetic evolutionary algorithm called BioIK that was developed during my research at the University of Hamburg.

The algorithm can be used for solving complex kinematic configurations on generic and fully-constrained geometries with multiple joint chains and objectives under high degree of freedom. It can overcome several limitations of traditional techniques, such as singularity issues and avoiding local minima solutions, while achieving similar computational speed and accuracy of numerical methods. Experiments are conducted on a large variety of different robot models, and applied to tasks like dexterous object manipulation, full-body motion generation, teleoperation, as well as collision-free trajectory generation, which is typically neglected for generic inverse kinematics solvers. I will discuss remaining limitations and preferred usecases of this method within the landscape of inverse kinematics problems, and point out some future directions and potential impact of this work at the end of the talk.

The Zoom meeting access information will be posted soon at the Informatics Department events website.

Monday, 20 August 2020, 10:00, Online-Meeting Hamburg

Speaker: Sebastian Starke, Informatics Dept., Univ. Hamburg