Register now After registration you will be able to apply for this opportunity online.
Exploring simple dexterous manipulation tasks with RL for hand design optimization
This project aims to develop a dexterous task using reinforcement learning (RL) within an IsaacGym-based environment, then optimize an anthropomorphic hand design to simplify its complexity while ensuring task success, ultimately building and testing the hand in a real-world setting.
Robotic hands can achieve dexterity through learning-based methods such as reinforcement learning (RL). However, the design of anthropomorphic hands require many DoFs and are complex and expensive to build. In this project, you will devise and implement a task requiring dexterity with a RL environment, and then incorporate it into our hand design optimization framework to simplify the design while still achieving the task. You will get to work with an IsaacGym-based RL environment, build the robot hand yourself based on our design architecture, and achieve the task in the real world.
**Requirements**:
- Willingness to prototype quickly and iterate through many versions
- Strong motivation, problem-solving skills, and ability to work independently
- Proficiency in RL / robotic simulators (e.g. IsaacGym, MuJoCo, stable-baselines)
Robotic hands can achieve dexterity through learning-based methods such as reinforcement learning (RL). However, the design of anthropomorphic hands require many DoFs and are complex and expensive to build. In this project, you will devise and implement a task requiring dexterity with a RL environment, and then incorporate it into our hand design optimization framework to simplify the design while still achieving the task. You will get to work with an IsaacGym-based RL environment, build the robot hand yourself based on our design architecture, and achieve the task in the real world.
**Requirements**: - Willingness to prototype quickly and iterate through many versions
- Strong motivation, problem-solving skills, and ability to work independently
**Work Packages**:
1. Review literature on RL and dexterous manipulation
2. Explore and propose dexterous tasks that can be feasibly implemented on a real-world platform
3. Implement RL tasks in IsaacGym (GPU simulation environment)
4. Optimize the hand design for the task using our framework
5. Implement task in the real life and run experiments
**Work Packages**:
1. Review literature on RL and dexterous manipulation
2. Explore and propose dexterous tasks that can be feasibly implemented on a real-world platform
3. Implement RL tasks in IsaacGym (GPU simulation environment)
4. Optimize the hand design for the task using our framework
5. Implement task in the real life and run experiments
To apply, please send Yasunori a short motivation statement for this project, with a copy of your CV, transcripts, and two reference contacts if you have worked on any past projects. Also feel free to contact me for any questions!
Yasunori Toshimitsu, PhD Candidate, ytoshimitsu@ethz.ch, Soft Robotics Lab, Institute of Robotics and Intelligent Systems, D-MAVT, ETH Zurich
Davide Liconti, PhD Student, dliconti@ethz.ch, Soft Robotics Lab, Institute of Robotics and Intelligent Systems, D-MAVT, ETH Zurich
Prof. Robert Katzschmann, Assistant Professor of Robotics, rkk@ethz.ch, Soft Robotics Lab, Institute of Robotics and Intelligent Systems, D-MAVT, ETH Zurich
To apply, please send Yasunori a short motivation statement for this project, with a copy of your CV, transcripts, and two reference contacts if you have worked on any past projects. Also feel free to contact me for any questions!
Yasunori Toshimitsu, PhD Candidate, ytoshimitsu@ethz.ch, Soft Robotics Lab, Institute of Robotics and Intelligent Systems, D-MAVT, ETH Zurich
Davide Liconti, PhD Student, dliconti@ethz.ch, Soft Robotics Lab, Institute of Robotics and Intelligent Systems, D-MAVT, ETH Zurich
Prof. Robert Katzschmann, Assistant Professor of Robotics, rkk@ethz.ch, Soft Robotics Lab, Institute of Robotics and Intelligent Systems, D-MAVT, ETH Zurich
