TU Berlin

Robotics and Biology LaboratoryUnderstanding Human Kinematic Problem Solving

Fachgebietslogo RBO


zur Navigation

Es gibt keine deutsche Übersetzung dieser Webseite.

Understanding Human Kinematic Problem Solving


Humans are exceptionally good at exploring, understanding and
manipulating kinematic structures such as articulated pieces of
furniture or kinematic puzzles. If we want to enable similar levels of
expertise for robots, it makes sense to study how humans explore and
interact with kinematics. This way we are going to get a deeper
understanding for how humans approach kinematic problems that may help
us build more capable robots in the future.

Description of Work

In this thesis you will use a modular kinematic puzzle (lockbox) to
design and perform experiments with human subjects. Using motion
tracking and other sensor measurements, you will record how the subjects
explore and try to solve different kinematic puzzle setups that you
designed. Based on that data, you will then use methods from data-analysis
like PCA or clustering to reveal and understand the subjects' strategies.
You will also implement candidate strategies for the behavior using
probabilistic methods, simulate these and optionally use these simulated
strategies in a bayesian experimental design approach to design new
lockbox experiments for human subjects to be tested.

Contact: Manuel Baum, Oliver Brock



Schnellnavigation zur Seite über Nummerneingabe