Smart manipulation

With recent advancements in artificial intelligence, smart manipulation has become an increasingly popular topic in industrial applications. We study the application of traditional computer vision techniques in combination with methods based on machine learning and neural networks for vision-based navigation and data acquisition tasks. We employ metaheuristics and approaches developed as a part of our planning-focused research for the inspection and motion planning tasks. Our research in the area of smart manipulation is focused on problems concerning uncertainties in the tasks of object pose estimation, manipulation, and visual inspection.

Another area of our research is focussed on smart materials and self-assembly in macroscopic-scale systems, where we cooperate with reaserchers from the Faculty of Civil Engineering at CTU. These assembling systems are being quite widely researched and used in biomedicine and nanotechnologies. We study the use of mechanical and magneto-mechanical shakers, as well as electromagnetic tile excitations, and address the limitations present in the macroscale world.

Object pose estimation and manipulation

We focus on complex problems with pose estimation of challenging objects, such as automotive parts with reflective properties, semi-transparent objects, or soft objects with partly varying shapes. We address the following tasks:

  • Object segmentation and pose estimation – a core problem in autonomous manipulation. This area of research concerns almost all tasks involving reasoning, manipulation, or interaction with objects.
  • Random bin picking – we develop complete solutions for the bin picking problem in which objects with random poses are picked out of a transportation bin by a robotic manipulator.
  • Force control - precise force control is imperative for safe interaction with objects, other robots, or even humans. It is also a necessity while working with dual-arm manipulators.



Publications

  • Kozák, V., Sushkov, R., Kulich, M., and Přeučil, L. (2021). Data-Driven Object Pose Estimation in a Practical Bin-Picking Application. Sensors.
    PDFURLBibTeX
    @Article{kozak2021sensorsBinpicking,
      author = {Kozák, Viktor and Sushkov, Roman and Kulich, Miroslav and Přeučil, Libor},
      title = {Data-Driven Object Pose Estimation in a Practical Bin-Picking Application},
      JOURNAL = {Sensors},
      VOLUME = {21},
      YEAR = {2021},
      NUMBER = {18},
      ARTICLE-NUMBER = {6093},
      URL = {https://www.mdpi.com/1424-8220/21/18/6093},
      ISSN = {1424-8220},
      DOI = {10.3390/s21186093}
    }
    
    
  • Surák, M., Košnar, K., Kulich, M., Kozák, V., and Přeučil, L. (2019). Visual Data Simulation for Deep Learning in Robot Manipulation Tasks. Modelling and Simulation for Autonomous Systems (MESAS 2018).
    PDFURLBibTeX
    @inproceedings{Surak19mesas,
      author = {Surák, M. and Košnar, Karel and Kulich, Miroslav and Kozák, Viktor and Přeučil, Libor},
      title = {Visual Data Simulation for Deep Learning in Robot Manipulation Tasks},
      booktitle = {Modelling and Simulation for Autonomous Systems (MESAS 2018)},
      publisher = {Springer International Publishing AG},
      address = {Cham, CH},
      year = {2019},
      language = {English},
      url = {https://doi.org/10.1007/978-3-030-14984-0_29},
      doi = {10.1007/978-3-030-14984-0_29},
      access = {accepted}
    }
    
    

Visual inspection

We address the following subproblems:

  • Navigation – we develop approaches for visual navigation and motion planning for robots and robotic manipulators during autonomous inspection procedures.
  • Classification and categorization – a core problem in visual inspection. In this area of research, we focus on the advancement of machine learning techniques for reasoning and knowledge extraction under uncertainties.
  • 3D reconstruction and model processing – precise spatial and semantic representation of data is a fundamental part of many applications. As such, we study and develop advanced techniques for data processing and 3D reconstruction.



Publications

  • Dörfler M., Pivoňka T., Košnar K., Přeucil L. (2020) Application of Surface Reconstruction for Car Undercarriage Inspection. 2020 3rd International Conference on Intelligent Robotic and Control Engineering (IRCE). IEEE, pp 47-51
    URLBibTeX
    @inproceedings{Dorfler2020SurfaceReconstruction,
    author="D{\"o}rfler, Martin
    and Pivo{\v{n}}ka, Tom{\'a}{\v{s}}
    and Ko{\v{s}}nar, Karel
    and P{\v{r}}eu{\v{c}}il, Libor",
    title="Application of Surface Reconstruction for Car Undercarriage Inspection",
    booktitle="Modelling and Simulation for Autonomous Systems",
    year="2020",
    publisher="IEEE",
    pages="47-51",
    doi="10.1109/IRCE50905.2020.9199251"
    }
    
    
  • Pivoňka, T., Košnar, K., Dörfler, M., Přeučil, L. (2019). Visual Odometry for Vehicles’ Undercarriage 3D Modelling. Modelling and Simulation for Autonomous Systems. MESAS 2018. Lecture Notes in Computer Science, vol 11472. Springer, Cham, pp 111-120
    PDFURLBibTeX
    @inproceedings{Pivonka2020ORB2TaR,
    author="Pivo{\v{n}}ka, Tom{\'a}{\v{s}}
    and Ko{\v{s}}nar, Karel
    and D{\"o}rfler, Martin
    and P{\v{r}}eu{\v{c}}il, Libor",
    editor="Mazal, Jan",
    title="Visual Odometry for Vehicles' Undercarriage 3D Modelling",
    booktitle="Modelling and Simulation for Autonomous Systems",
    year="2019",
    publisher="Springer International Publishing",
    address="Cham",
    pages="111--120"
    }
    
    

Contact: Libor Přeučil

Page last modified on September 08, 2022, at 09:48 AM EST
Powered by PmWiki