Register now After registration you will be able to apply for this opportunity online.
GPU Acceleration of Soft Robot Modeling: Enhancing Performance with CUDA
We are enhancing soft robot modeling by developing a GPU-accelerated version of our FEM-based framework using CUDA. This research focuses on optimizing parallel computations to significantly speed up simulations, enabling larger problem sizes and real-time control. By improving computational efficiency, we aim to advance soft robotics research and facilitate more detailed, dynamic simulations.
Keywords: Soft Body Simulation, high-performance computing, GPU programming, Parallel Computing, Finite Element Method (FEM), Multiphysics Simulation
The Soft Robotics Lab is developing a framework for soft robot modeling using the Finite Element Method (FEM) and an energy minimization model. The current implementation is optimized for parallel execution on CPUs. To meet the demand for more complex simulations and improve computational efficiency, we propose developing a GPU-accelerated version of the framework. By harnessing the computational power of GPUs, we aim to significantly speed up simulations and scale up problem sizes, enabling more detailed and realistic soft robotics simulations as well as real-time control. GPUs excel in handling numerous parallel computations simultaneously, making them ideal for speeding up iterative calculations
inherent in soft body dynamics simulations. Ultimately, this effort contributes to pushing the boundaries of soft robotics research and application, paving the way for innovative developments in the field.
The Soft Robotics Lab is developing a framework for soft robot modeling using the Finite Element Method (FEM) and an energy minimization model. The current implementation is optimized for parallel execution on CPUs. To meet the demand for more complex simulations and improve computational efficiency, we propose developing a GPU-accelerated version of the framework. By harnessing the computational power of GPUs, we aim to significantly speed up simulations and scale up problem sizes, enabling more detailed and realistic soft robotics simulations as well as real-time control. GPUs excel in handling numerous parallel computations simultaneously, making them ideal for speeding up iterative calculations inherent in soft body dynamics simulations. Ultimately, this effort contributes to pushing the boundaries of soft robotics research and application, paving the way for innovative developments in the field.
1. Understand the current soft body modeling framework and its CPU parallelization.
2. Review literature on CUDA and GPU acceleration for FEM.
3. Set up a development environment for GPU programming.
4. Fine-tune CUDA kernels and parallelization strategies for maximum performance.
5. Conduct benchmark simulations to verify correctness and performance gains.
1. Understand the current soft body modeling framework and its CPU parallelization. 2. Review literature on CUDA and GPU acceleration for FEM. 3. Set up a development environment for GPU programming. 4. Fine-tune CUDA kernels and parallelization strategies for maximum performance. 5. Conduct benchmark simulations to verify correctness and performance gains.
- Strong academic background with exceptional grades in computer science.
- Strong programming skills (C++ and Python).
- Knowledge in FEM-based simulations.
- Knowledge in basic GPU Programming including CUDA.
- Keen to learn more about soft robotics and physical modeling.
- Capable of both working independently and cooperating with mentors and teammates.
- Strong academic background with exceptional grades in computer science. - Strong programming skills (C++ and Python). - Knowledge in FEM-based simulations. - Knowledge in basic GPU Programming including CUDA. - Keen to learn more about soft robotics and physical modeling. - Capable of both working independently and cooperating with mentors and teammates.
Manuel Mekkattu, manuel.mekkattu@srl.ethz.ch, Soft Robotics Lab, D-MAVT, ETH Zürich
Prof. Robert Katzschmann, rkk@ethz.ch, Soft Robotics Lab, D-MAVT, ETH Zürich
Earliest start: February 1, 2025
If you are interested, please submit a motivation letter, your CV, and transcripts.
Manuel Mekkattu, manuel.mekkattu@srl.ethz.ch, Soft Robotics Lab, D-MAVT, ETH Zürich
Prof. Robert Katzschmann, rkk@ethz.ch, Soft Robotics Lab, D-MAVT, ETH Zürich
Earliest start: February 1, 2025
If you are interested, please submit a motivation letter, your CV, and transcripts.