Dr Marek Kopicki

Research interest

The simple task of grasping an object and packing it in a container is effortlessly done by a human, but it can be extremely challenging for a robot. This is a consequence of the complexity of possible variations of object appearance, shape, friction, mass, etc., and of the complexity of the changing relations between the robot's body and other objects in the environment.

I am interested in mechanisms which would enable a robot to autonomously perceive, act, learn and adapt to the environment using its vision, haptics and motor system. In particular, in mechanisms enabling perception and learning, the physical properties of objects from demonstration or through direct interaction, such as pushing, poking or grasping. Finding these mechanisms requires integrating ideas from a range of research fields such as computer vision, machine learning, motor control, planning and experimental psychology.

Most of my current research is focused on our dexterous grasp learning algorithm [Kopicki-et-all-2015][Kopicki-et-all-2014][Kopicki-2013 implemented within the Golem and Grasp robot control and learning frameworks. Early versions of the algorithm in action can be found [here], while the latest [here]. The latest version is also central to the robot Boris, which has recently received a lot of attention in the media [PaCMan-News].

Collaborations

Current research projects

• I am Co-Pi on RoMaNs: Robotic Manipulation for Nuclear Sort and Segregation
  • Ales Leonardis, Rustam Stolkin: visual representations for object grasping.
  • Gerhard Neumann, Jan Peters: reinforcement learning in grasp learning and planning.
• I am work-package leader on PaCMan: Probabilistic and Compositional Representations for Object Manipulation
  • Jeremy Wyatt, Renaud Detry: dexterous grasp learning.
  • Ales Leonardis, Krzysztof Walas: 3D compositional representations of object contacts.
  • Carlos Rosales Gallegos, Hamal Marino, Sebastian Zurek, Maxime Adjigble, Marco Gabiccini: grasp learning with Pisa Soft Hand.
  • Jeremy Wyatt, Caludio Zito: robotic grasping under uncertainty.
• I am Co-Pi on Machine-learning of vision-guided bi-manual grasps, for adaptable autonomous manipulation in manufacturing environments, sponsored by Cranfield-Loughborough EPSRC.
• I am support academic on a 3 years project with KUKA Robotics Company to develop a new toolbox of software and algorithms, which would extend KUKA’s robotics capabilities to include advanced computer vision, and vision-based robot-control.

Publications

Patents

• [kopicki2013patent] Marek Kopicki. Grasp Modelling. International Patent WO/2014/188177. 2013. [bib] [publisher]

Journals

• [kopicki2015a] Marek Kopicki, Renaud Detry, Maxime Adjigble, Rustam Stolkin, Ales Leonardis and Jeremy L. Wyatt. One shot learning and generation of dexterous grasps for novel objects. In The International Journal of Robotics Research. 2015. [pdf] [bib] [doi] [publisher] [videos]
• [kopicki2016learning] Marek Kopicki, Sebastian Zurek, Rustam Stolkin, Thomas Morwald and Jeremy L. Wyatt. Learning modular and transferable forward models of the motions of push manipulated objects. In Autonomous Robots. 2016. [pdf] [bib] [doi] [publisher]

Refereed Conference Publications

• [kopicki2014a] Marek Kopicki, Renaud Detry, Florian Schmidt, Christoph Borst, Rustam Stolkin and Jeremy L. Wyatt. Learning Dexterous Grasps That Generalise To Novel Objects By Combining Hand And Contact Models. In Proceedings of IEEE International Conference on Robotics and Automation (ICRA), pages 5358-5365. 2014. [pdf] [bib] [publisher] [videos]
• [belter2014kinematically] Dominik Belter, Marek Kopicki, Sebastian Zurek and Jeremy Wyatt. Kinematically optimised predictions of object motion. In Proceedings of IEEE International Conference on Intelligent Robots and Systems (ICRA), pages 4422-4427. 2014. [pdf] [bib] [publisher]
• [zito2013sequential] Claudio Zito, Marek Kopicki, Rustam Stolkin, Christoph Borst, Florian Schmidt, Maximo A. Roa and Jeremy Wyatt. Sequential trajectory re-planning with tactile information gain for dexterous grasping under object-pose uncertainty. In Proceedings of IEEE International Conference on Intelligent Robotics and Systems (IROS), pages 4013-4020. 2013. [pdf] [bib] [publisher]
• [zito2012two] Claudio Zito, Rustam Stolkin, Marek Kopicki and Jeremy L. Wyatt. Two-level RRT planning for robotic push manipulation. In Proceedings of IEEE International Conference on Intelligent Robotics and Systems (IROS), pages 678-685. 2012. [pdf] [bib] [publisher]
• [kopicki2011learning] Marek Kopicki, Sebastian Zurek, Rustam Stolkin, Thomas Mörwald and Jeremy L. Wyatt. Learning to predict how rigid objects behave under simple manipulation. In Proceedings of IEEE International Conference on Robotics and Automation (ICRA), pages 5722-5729. 2011. [pdf] [bib] [publisher]
• [morwald2011predicting] Thomas Mörwald, Marek Kopicki, Rustam Stolkin, Jeremy Wyatt, Sebastian Zurek, Michael Zillich and Markus Vincze. Predicting the unobservable Visual 3D tracking with a probabilistic motion model. In Proceedings of IEEE International Conference on Robotics and Automation (ICRA), pages 1849-1855. 2011. [pdf] [bib] [publisher]
• [kopicki2009prediction] Marek Kopicki, Jeremy L. Wyatt and Rustam Stolkin. Prediction learning in robotic pushing manipulation. In Proceedings of IEEE International Conference on Advanced Robotics (ICAR), pages 1-6. 2009. [pdf] [bib] [publisher]

Workshops

• [rietzler2013inertially] Alexander Rietzler, Renaud Detry, Marek Kopicki, Jeremy L. Wyatt and Justus Piater. Inertially-safe Grasping of Novel Objects. In IROS Workshop: Cognitive Robotics Systems: Replicating Human Actions and Activities. 2013. [pdf] [bib]
• [zito2013sequentialb] Claudio Zito, Marek Kopicki, Rustam Stolkin, Christoph Borst, Florian Schmidt, Maximo Roa and Jeremy L. Wyatt. Sequential Re-planning for Dextrous Grasping Under Object-pose Uncertainty. In Robotics: Science and Systems (RSS) Workshop: Manipulation with Uncertain Models. 2013. [pdf] [bib]
• [zito2012exploratory] Claudio Zito, Rustam Stolkin, Marek Kopicki, Massimiliano Di Luca and Jeremy L. Wyatt. Exploratory reach-to-grasp trajectories for uncertain object poses. In IROS Workshop: Beyond Robot Grasping: Modern Approaches for Dynamic Manipulation. 2012. [pdf] [bib]
• [kopicki2010predicting] Marek Kopicki, Rustam Stolkin, Sebastian Zurek, Thomas Mörwald and Jeremy L. Wyatt. Predicting workpiece motions under pushing manipulations using the principle of minimum energy. In Robotics: Science and Systems (RSS) Workshop: Representations for object grasping and manipulation in single and dual arm tasks. 2010. [pdf] [bib]

Thesis

• [kopicki2010prediction] Marek Kopicki. Prediction learning in robotic manipulation. Ph.D. thesis, Computer Science, University of Birmingham. 2010. [pdf] [bib] [library]

Book chapters

• [hawes2010playmate] Nick Hawes, Jeremy L. Wyatt, Mohan Sridharan, Marek Kopicki, Somboon Hongeng, Ian Calvert, Aaron Sloman, Geert-Jan Kruijff, Henrik Jacobsson, Michael Brenner, Danijel Skocaj, Alen Vrecko, Nikodem Majer and Michael Zillich. The PlayMate System. In Henrik I. Christensen and Geert-Jan M. Kruijff and Jeremy L. Wyatt (editors)Cognitive Systems, volume 8 of Cognitive Systems Monographs, pages 367—393, Springer Berlin Heidelberg. 2010. [bib] [publisher]

Reports

• [kopicki2005report3] Marek Kopicki. Learning object affordances by imitation. Technical Report, Computer Science, University of Birmingham. 2005. [pdf] [bib]
• [kopicki2004project] Marek Kopicki. Robotic mapping. Technical Report, Computer Science, University of Birmingham. 2004. [pdf] [bib]
• [kopicki2004report2] Marek Kopicki. Monte Carlo Localisation for mobile robots. Technical Report, Computer Science, University of Birmingham. 2004. [pdf] [bib]

Software

Golem and Grasp are C++ robot control, planning and manipulation learning frameworks are available on [my GitHub website].