Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Poke it: Towards deformable object manipulation using poking strategies.
Despite recent advances in object manipulation using real-world data [1], [2], deformable object manipulation remains an unexplored field as most of the efforts have mainly focused their attention on rigid objects. Other approaches, leveraging simulation and learning [3], [4], have shown success in grasping tasks for deformable objects, however, their applicability to manipulation tasks has not been validated and they rely on accurate estimation of the material properties to bridge the sim-to-real gap. The core of this project is deformable object manipulation and material estimation based on force feedback during interaction.
The student will build a robust framework for poking deformable objects, leveraging our existing C++ code base that already enables robot control and teleoperation of different platforms (YuMi, UR5, Franka Panda, Kinova). Depending on the scope of the thesis, the student will also implement learning-based approaches to estimate material properties (stiffness, friction coefficients, etc.) or efficient Imitation Learning [1], [2] or Reinforcement Learning [5] algorithms to train goal-oriented policies with real-world data. A literature review will be part of the thesis and a written report and an oral presentation will conclude it.
The student will build a robust framework for poking deformable objects, leveraging our existing C++ code base that already enables robot control and teleoperation of different platforms (YuMi, UR5, Franka Panda, Kinova). Depending on the scope of the thesis, the student will also implement learning-based approaches to estimate material properties (stiffness, friction coefficients, etc.) or efficient Imitation Learning [1], [2] or Reinforcement Learning [5] algorithms to train goal-oriented policies with real-world data. A literature review will be part of the thesis and a written report and an oral presentation will conclude it.
Depending on the scope of the thesis, this project has two goals, first to build a robust data collection pipeline for poking real-world deformable objects, and second to explore Imitation Learning approaches [2], [1] or data-efficient Reinforcement Learning approaches (online or offline) [5] to train goal-oriented policies.
Depending on the scope of the thesis, this project has two goals, first to build a robust data collection pipeline for poking real-world deformable objects, and second to explore Imitation Learning approaches [2], [1] or data-efficient Reinforcement Learning approaches (online or offline) [5] to train goal-oriented policies.
Hehui Zheng (hehui.zheng@srl.ethz.ch) or Miguel Zamora (miguel.zamora@inf.ethz.ch)
Hehui Zheng (hehui.zheng@srl.ethz.ch) or Miguel Zamora (miguel.zamora@inf.ethz.ch)