Computational Design and Optimization for Physically Intelligent Systems
Event details
| Date | 29.05.2026 |
| Hour | 13:00 › 15:00 |
| Speaker | Junyu Liu |
| Location | |
| Category | Conferences - Seminars |
EDIC candidacy exam
Exam president: Prof. Wenzel Jakob
Thesis advisor: Prof. Mark Pauly
Co-examiner: Prof. Pascal Fua
Abstract
Physically intelligent systems exploit geometry, physics, material distribution, and structural coupling to generate useful behaviors. This thesis develops computational design and optimization methods for lightweight mechanical systems with complex geometries and rich deformation behavior. The central goal is to create a simulation-driven inverse design framework in which shape, material, and structural parameters are optimized together so that the resulting physical system achieves desired equilibrium configurations, alignment targets, and crucially functional mechanical responses. By combining differentiable mechanics, compact design parameterizations, robust equilibrium solving, instructive shape space and functional space analysis, and optimization, the work investigates how complex physical behavior can be encoded directly into the design process.
Selected papers
- Computational Design and Automated Fabrication of Kirchhoff-Plateau Surfaces: https://dl.acm.org/doi/pdf/10.1145/3072959.3073695
- The Design Space of Plane Elastic Curves: https://cdl.ethz.ch/publications/the-design-space-of-plane-elastic-curves/
- Procedural Metamaterials: A Unified Procedural Graph for Metamaterial Design: https://cdl.ethz.ch/publications/procedural-metamaterials-a-unified-procedural-graph-for-metamaterial-design/
Exam president: Prof. Wenzel Jakob
Thesis advisor: Prof. Mark Pauly
Co-examiner: Prof. Pascal Fua
Abstract
Physically intelligent systems exploit geometry, physics, material distribution, and structural coupling to generate useful behaviors. This thesis develops computational design and optimization methods for lightweight mechanical systems with complex geometries and rich deformation behavior. The central goal is to create a simulation-driven inverse design framework in which shape, material, and structural parameters are optimized together so that the resulting physical system achieves desired equilibrium configurations, alignment targets, and crucially functional mechanical responses. By combining differentiable mechanics, compact design parameterizations, robust equilibrium solving, instructive shape space and functional space analysis, and optimization, the work investigates how complex physical behavior can be encoded directly into the design process.
Selected papers
- Computational Design and Automated Fabrication of Kirchhoff-Plateau Surfaces: https://dl.acm.org/doi/pdf/10.1145/3072959.3073695
- The Design Space of Plane Elastic Curves: https://cdl.ethz.ch/publications/the-design-space-of-plane-elastic-curves/
- Procedural Metamaterials: A Unified Procedural Graph for Metamaterial Design: https://cdl.ethz.ch/publications/procedural-metamaterials-a-unified-procedural-graph-for-metamaterial-design/
Practical information
- General public
- Free