Maxim Vochten

Department of Mechanical Engineering,

Celestijnenlaan 300,

3001 Leuven, Belgium.

About

I am a Postdoctoral Research Associate within the Robotics Research Group at KU Leuven.

My main aim in research is to improve the collaboration between humans and robots by formulating data-efficient task representations that can be used for human intent recognition and for the fast programming of reactive robot motions. To achieve this, I combine concepts from numerical optimization, optimal control, differential geometry, screw theory, 3D kinematics, state estimation and robot control.

I am currently working on the Robotgenskill project which focuses on the generalization of human-demonstrated robot skills.

Students

I am the co-promotor of two PhD candidates:

Arno Verduyn. Topic: Analysis, recognition and generalization of human-demonstrated skills using invariant trajectory representations
Riccardo Burlizzi. Topic: Data augmentation in robot learning from demonstration using invariant trajectory representations

I have mentored and co-supervised over 15 master thesis students in the Master of Mechanical Engineering programme (and the Master of Artificial Intelligence programme) at KU Leuven.

Software

For my research on invariant trajectory descriptors, I am developing software packages in Matlab and Python: invariants_mat and invariants_py. These packages support the robust numerical calculation of invariant descriptors and also support the generation of trajectories in an invariant way. A corresponding documentation website is being developed: Trajectory Invariants.

In my research I frequently make use of the following specialized software libraries for robotics:

eTaSL for task specification and reactive sensor-based control
CasADi for solving optimal control problems and nonlinear optimization problems
ROS for building robotics applications integrating sensing, motion generation, control, and visualization
Orocos for building fast robot controllers

News

Apr 18, 2024	At the upcoming ICRA 2024 conference in Yokohama I will present our work on invariant trajectory descriptors and Arno will present our work on self-supervised trajectory segmentation. Presentations are on Wednesday 15th of May.
Jan 19, 2024	Our paper on self-supervised trajectory segmentation has been accepted for the ICRA 2024 conference in Yokohama, Japan!
Dec 20, 2023	Our paper on interpreting contact tasks using coordinate-invariant trajectory representations has been published in the December 2023 issue of the IEEE Transactions on Robotics.
Aug 21, 2023	Arno Verduyn will present our paper on trajectory generation for robotic spray painting at the upcoming IEEE CASE conference in Auckland, New Zealand, on 28 August 2023.

Recent publications

TRO

Invariant Descriptors of Motion and Force Trajectories for Interpreting Object Manipulation Tasks in Contact

Maxim Vochten, Ali Mousavi Mohammadi, Arno Verduyn, Tinne De Laet, Erwin Aertbeliën, and Joris De Schutter

IEEE Transactions on Robotics, 2023

Bib HTML

@article{vochten2023invariant,
  author = {Vochten, Maxim and Mohammadi, Ali Mousavi and Verduyn, Arno and De Laet, Tinne and Aertbeliën, Erwin and De Schutter, Joris},
  journal = {IEEE Transactions on Robotics},
  title = {Invariant Descriptors of Motion and Force Trajectories for Interpreting Object Manipulation Tasks in Contact},
  year = {2023},
  volume = {39},
  number = {6},
  pages = {1-21},
  issn = {1941-0468},
  publisher = {Institute of Electrical and Electronics Engineers (IEEE)},
  url = {http://dx.doi.org/10.1109/TRO.2023.3309230},
  doi = {10.1109/TRO.2023.3309230},
}