08-17, 14:40–15:00 (Europe/Zurich), HS 120
Rigid transformation in 3D are complicated due to the multitude of different conventions and because they often form complex graphs that are difficult to manage. In this talk I will give a brief introduction to the topic and present the library pytransform3d as a set of tools that can help you to tame the complexity. Throughout the talk I will use examples from robotics (imitation learning, collision detection, state estimation, kinematics) to motivate the discussed features, even though presented solutions are useful beyond robotics.
This talk focuses on rotation and translation, that is, rigid transformations, in three dimensions. There are various representations of these. We often combine several software components with different conventions. Furthermore, we usually combine multiple transformations that form complex graphs of transformations, and we are often interested in transformations that are not directly available, but can be computed from a combination of multiple transformations. Both problems can be handled with pytransform3d, a Python library for transformations in three dimensions.
pytransform3d offers...
- operations for most common representations of rotation / orientation and
translation / position - conversions between those representations
- clear documentation of conventions
- tight coupling with matplotlib to quickly visualize (or animate)
transformations - the TransformManager which organizes complex chains of transformations
- the UrdfTransformManager which is able to load transformations from URDF
files - a matplotlib-like interface to Open3D’s visualizer to display geometries and
transformations
I will present several features of the library in this talk and I will use examples from robotics for illustration, for example,
- imitation learning - learning robotic motion from human demonstration
- kinematics - translation of a human hand motion to a robotic hand
- collision detection - between a robot arm and it's environment
- state estimation - estimation of a robot's location and its uncertainty
There are several pitfalls that we will discuss as well.
Robotics & IoT
Expected audience expertise: Domain –none
Expected audience expertise: Python –some
Project Homepage / Git – Abstract as a tweet –Introduction to 3D transformations in Python illustrated with problems in robotics
Alexander Fabisch received his diploma degree in computer science from the University of Bremen in 2012. From 2012 to 2017 he worked as a researcher at the robotics research group of the University of Bremen and since 2017 he works at the Robotics Innovation Center of the German Research Center for Artificial Intelligence (DFKI). He obtained his doctoral degree from University of Bremen in 2020. His scientific interests are in the fields of machine learning and black-box optimization with robotic applications and a focus on learning manipulation behaviors.