Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Data-driven Control of a Muscle-Actuated Mechanical Hand
Design, build, and implement the control of a multi-DOF robot joint actuated by multiple pneumatic artificial muscles (PAMs). We will explore behaviors that can be produced by over-actuating a sequence of multi-dof joints with muscle-actuated tendons, and investigate how levels of coupling between actuated degrees of freedom can be used to simplify actuation.
Our finger digits are controlled by muscles in our forearm connected through tendons allowing for a high-DOF motion in our fingers. In addition to that, the antagonistic muscle configuration results in a level of versatility and dexterity that remains difficult to replicate in a mechanical design.
In this project, we aim to explore how coupled, over-actuated systems can be used to achieve levels of dexterity and versatility such as in the human hand. The student shall 1) build an overactuated multi-dof joint actuated by PAMs, 2) map pressure values in the PAMs to the tilting and twisting motions of the effector 3) investigate using machine learning assignment of PAMs to a smaller subset of pressure inputs, thereby introducing coupling between the actuators while keeping dof of motion high.
**Work Packages**
1. Literature review of existing work on modeling of PAMs, data-driven control, etc.
2. Build a single overactuated multi-DOF joint driven by PAMs
3. Data-driven modeling and control of the system
4. Investigate introducing coupling between actuators
5. Design and control of a multi joint setup, e.g., finger
**Requirements**
1. High motivation and problem-solving ability
2. Capable of both working independently and cooperating in a team
3. Keen to learn more about soft robotics and robot control
4. Proficiency in C/C++ and/or Python
5. Knowledge of Newtonian mechanics and basic continuum mechanics
6. Experience in rapid-prototyping, CAD, and/or robot simulation is a plus
Our finger digits are controlled by muscles in our forearm connected through tendons allowing for a high-DOF motion in our fingers. In addition to that, the antagonistic muscle configuration results in a level of versatility and dexterity that remains difficult to replicate in a mechanical design.
In this project, we aim to explore how coupled, over-actuated systems can be used to achieve levels of dexterity and versatility such as in the human hand. The student shall 1) build an overactuated multi-dof joint actuated by PAMs, 2) map pressure values in the PAMs to the tilting and twisting motions of the effector 3) investigate using machine learning assignment of PAMs to a smaller subset of pressure inputs, thereby introducing coupling between the actuators while keeping dof of motion high.
**Work Packages** 1. Literature review of existing work on modeling of PAMs, data-driven control, etc. 2. Build a single overactuated multi-DOF joint driven by PAMs 3. Data-driven modeling and control of the system 4. Investigate introducing coupling between actuators 5. Design and control of a multi joint setup, e.g., finger
**Requirements** 1. High motivation and problem-solving ability 2. Capable of both working independently and cooperating in a team 3. Keen to learn more about soft robotics and robot control 4. Proficiency in C/C++ and/or Python 5. Knowledge of Newtonian mechanics and basic continuum mechanics 6. Experience in rapid-prototyping, CAD, and/or robot simulation is a plus
In this project, we aim to explore how coupled, over-actuated systems can be used to achieve levels of dexterity and versatility such as in the human hand. The student shall 1) build an overactuated multi-dof joint actuated by PAMs, 2) map pressure values in the PAMs to the tilting and twisting motions of the effector 3) investigate using machine learning assignment of PAMs to a smaller subset of pressure inputs, thereby introducing coupling between the actuators while keeping dof of motion high.
In this project, we aim to explore how coupled, over-actuated systems can be used to achieve levels of dexterity and versatility such as in the human hand. The student shall 1) build an overactuated multi-dof joint actuated by PAMs, 2) map pressure values in the PAMs to the tilting and twisting motions of the effector 3) investigate using machine learning assignment of PAMs to a smaller subset of pressure inputs, thereby introducing coupling between the actuators while keeping dof of motion high.
- Barnabas Gavin Cangan, gavin,cangan@srl.ethz.ch, Soft Robotics Lab, D-MAVT Thomas Buchner, thomas.buchner@srl.ethz.ch, Soft Robotics Lab, D-MAVT
- Dr. Espen Knoop, espen.knoop@disneyresearch.com, Disney Research
- Dr. Moritz Bächer, moritz.baecher@disneyresearch.com, Disney Research
- Prof. Dr. Robert Katzschmann, rkk@ethz.ch, Soft Robotics Lab, D-MAVT
Please submit **your CV, your BSc and MSc transcripts, and two reference contacts.** Applications are accepted via SiROP.
Please visit for all available student projects: https://srl.ethz.ch/education/student-projects.html
- Barnabas Gavin Cangan, gavin,cangan@srl.ethz.ch, Soft Robotics Lab, D-MAVT Thomas Buchner, thomas.buchner@srl.ethz.ch, Soft Robotics Lab, D-MAVT - Dr. Espen Knoop, espen.knoop@disneyresearch.com, Disney Research - Dr. Moritz Bächer, moritz.baecher@disneyresearch.com, Disney Research - Prof. Dr. Robert Katzschmann, rkk@ethz.ch, Soft Robotics Lab, D-MAVT
Please submit **your CV, your BSc and MSc transcripts, and two reference contacts.** Applications are accepted via SiROP. Please visit for all available student projects: https://srl.ethz.ch/education/student-projects.html