Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Computational Modeling of Biological Muscle for Sim-to-Real Design of Muscle Actuators for Biohybrid Robots
One of the most significant limitations in advancing the field of biohybrid robotics is the extensive trial-and-error approaches necessary to engineer biological muscle for actuating a biohybrid robot. There is no accurate computational model to describe the contraction behavior of a muscle construct. In this cutting edge project at the intersection of computational modeling and biology, we aim to develop a computational model of biological muscles for sim-to-real applications.
Our starting point will be a differentiable simulation framework that has been developed for computational design of soft body swimmers, where the muscle is approximated as a Neo-Hookean material with a spring-like actuation. We will also implement a different muscle model based on Cosserat rod theory (https://www.cosseratrods.org/). We will run comparisons of results between the two models, and optimize the models to accurately recapitulate the contraction of muscle constructs engineered in the lab. The final goal will be to validate the models on experimental data and to better understand the necessary features of a sim-to-real biological muscle model. Your final thesis will be a simulation-based work allowing you to further develop your computational modeling skills and gain real-world experience on the techniques and challenges of matching computational models to experimental data.
Please see attached pdf for further details & selected references.
**Work Packages**
1. Literature review on current methods to model biological muscle
2. Implementation of muscle model in differentiable simulation & Cosserat rod frameworks
3. Comparison between models & optimization to recapitulate muscle contraction
4. Experimental validation of models and comparison between models for sim-to-real application
**Requirements**
- Background in computer sciences or engineering
- High motivation and problem-solving ability, desire to learn and apply new techniques
- Capability and desire to work independently
- Strong programming skills: Python, C++
- Experience with soft body, multiphysics, and/or differentiable simulations, and machine learning applied to simulations, is a plus
Our starting point will be a differentiable simulation framework that has been developed for computational design of soft body swimmers, where the muscle is approximated as a Neo-Hookean material with a spring-like actuation. We will also implement a different muscle model based on Cosserat rod theory (https://www.cosseratrods.org/). We will run comparisons of results between the two models, and optimize the models to accurately recapitulate the contraction of muscle constructs engineered in the lab. The final goal will be to validate the models on experimental data and to better understand the necessary features of a sim-to-real biological muscle model. Your final thesis will be a simulation-based work allowing you to further develop your computational modeling skills and gain real-world experience on the techniques and challenges of matching computational models to experimental data.
Please see attached pdf for further details & selected references.
**Work Packages**
1. Literature review on current methods to model biological muscle
2. Implementation of muscle model in differentiable simulation & Cosserat rod frameworks
3. Comparison between models & optimization to recapitulate muscle contraction
4. Experimental validation of models and comparison between models for sim-to-real application
**Requirements**
- Background in computer sciences or engineering
- High motivation and problem-solving ability, desire to learn and apply new techniques
- Capability and desire to work independently
- Strong programming skills: Python, C++
- Experience with soft body, multiphysics, and/or differentiable simulations, and machine learning applied to simulations, is a plus
An experimentally validated computational model of biological muscle for sim-to-real application.
An experimentally validated computational model of biological muscle for sim-to-real application.
Aiste Balciunaite, abalciunaite@ethz.ch, Institute of Robotics and Intelligent Systems, D-MAVT
Dr. Öncay Yasa, yasao@ethz.ch, Institute of Robotics and Intelligent Systems, D-MAVT
Prof. Robert Katzschmann, rkk@ethz.ch, Institute of Robotics and Intelligent Systems, D-MAVT
Aiste Balciunaite, abalciunaite@ethz.ch, Institute of Robotics and Intelligent Systems, D-MAVT
Dr. Öncay Yasa, yasao@ethz.ch, Institute of Robotics and Intelligent Systems, D-MAVT
Prof. Robert Katzschmann, rkk@ethz.ch, Institute of Robotics and Intelligent Systems, D-MAVT