Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Large Scale Stone Aggregation
This project is about autonomous object aggregation with a walking excavator. The goal is to explore, implement, and evaluate methods for discrete object assembly on a full scale construction machine.
Keywords: Discrete Assembly, Autonomous Construction, 3D Reconstruction and Segmentation, Manipulation, Force Control
We at RSL and ASL want to perform autonomous and robust robotic assembly of objects with irregular geometry and uncertainties in pose and shape. This project is part of our effort to bring environment aware manipulation skills to a large scale construction machine to further push the boundaries of automated on site construction.
In our previous work we developed a system able to autonomously construct stable structures with found irregular stones [1]. During this work we already created a set of skills and tools for manipulation with irregularly shaped objects, like object localization, segmentation, and reconstruction or grasping pose generation and evaluation. The goal of this project is to transfer and adapt these methods for manipulation with a walking excavator [2]. This full scale construction machine is equipped with force controllable actuators and sensors for localization and mapping, making it a versatile robotic platform for the use in unstructured environments. Such a platform will for example allow to perform automated landscaping and river bed construction with large stone blocks or to reuse concrete rubble directly on a construction site.
[1] Furrer, Fadri, Martin Wermelinger, Hironori Yoshida, Fabio Gramazio, Matthias Kohler, Roland Siegwart, and Marco Hutter. "Autonomous Robotic Stone Stacking with Online next Best Object Target Pose Planning." IEEE International Conference on Robotics and Automation (ICRA), 2017.
[2] Hutter, Marco, Philipp Leemann, Stefan Stevsic, Andreas Michel, Dominic Jud, Mark Hoepflinger, Roland Siegwart et al. "Towards optimal force distribution for walking excavators." In International Conference on Advanced Robotics (ICAR), 2015.
We at RSL and ASL want to perform autonomous and robust robotic assembly of objects with irregular geometry and uncertainties in pose and shape. This project is part of our effort to bring environment aware manipulation skills to a large scale construction machine to further push the boundaries of automated on site construction.
In our previous work we developed a system able to autonomously construct stable structures with found irregular stones [1]. During this work we already created a set of skills and tools for manipulation with irregularly shaped objects, like object localization, segmentation, and reconstruction or grasping pose generation and evaluation. The goal of this project is to transfer and adapt these methods for manipulation with a walking excavator [2]. This full scale construction machine is equipped with force controllable actuators and sensors for localization and mapping, making it a versatile robotic platform for the use in unstructured environments. Such a platform will for example allow to perform automated landscaping and river bed construction with large stone blocks or to reuse concrete rubble directly on a construction site.
[1] Furrer, Fadri, Martin Wermelinger, Hironori Yoshida, Fabio Gramazio, Matthias Kohler, Roland Siegwart, and Marco Hutter. "Autonomous Robotic Stone Stacking with Online next Best Object Target Pose Planning." IEEE International Conference on Robotics and Automation (ICRA), 2017.
[2] Hutter, Marco, Philipp Leemann, Stefan Stevsic, Andreas Michel, Dominic Jud, Mark Hoepflinger, Roland Siegwart et al. "Towards optimal force distribution for walking excavators." In International Conference on Advanced Robotics (ICAR), 2015.
Depending on the interests and strengths of the candidate, we can offer the following work packages:
- 3D detection, localization, segmentation, and reconstruction of discrete single objects.
- Evaluation/Design of appropriate gripping tool for stones and rubble waste.
- Grasping pose detection and selection for irregularly shaped objects.
- Trajectory planning and optimization.
- Development and implementation of method to detect and control the interaction of the object with the environment during placing under uncertainties.
- Evaluation and demonstration of the methods in an existing simulation environment and on the real robot.
Depending on the interests and strengths of the candidate, we can offer the following work packages: - 3D detection, localization, segmentation, and reconstruction of discrete single objects. - Evaluation/Design of appropriate gripping tool for stones and rubble waste. - Grasping pose detection and selection for irregularly shaped objects. - Trajectory planning and optimization. - Development and implementation of method to detect and control the interaction of the object with the environment during placing under uncertainties. - Evaluation and demonstration of the methods in an existing simulation environment and on the real robot.
We seek a highly motivated student with skills in C++ programming. Experience with ROS is beneficial.
We seek a highly motivated student with skills in C++ programming. Experience with ROS is beneficial.
Please contact Martin Wermelinger (martin.wermelinger@mavt.ethz.ch) for any questions. Your application should include a very brief motivational statement, your transcript of records and your CV.
Please contact Martin Wermelinger (martin.wermelinger@mavt.ethz.ch) for any questions. Your application should include a very brief motivational statement, your transcript of records and your CV.