Presentation
Differentiable Physics for Graphics and AI
DescriptionDifferentiable Physics for Graphics and AI (DPGA) is a half-day SIGGRAPH technical workshop exploring how differentiable physics-based simulation is transforming graphics and related fields including robotics, 3D vision, design, and fabrication. As simulation becomes tightly integrated with optimization, learning, and generative pipelines, differentiability enables new workflows for inverse problems, system identification, and physics-aware design.
The workshop is organized by researchers from Carnegie Mellon University, UCLA, the University of Utah, and Meta Reality Labs, whose work spans physical simulation, graphics systems, robotics, and AI. Together with invited speakers from academia and industry, they bring perspectives from foundational simulation research to large-scale production systems and real-world applications.
The workshop brings together foundational methods, practical systems, and emerging applications in differentiable simulation across graphics, robotics, 3D vision, design, and fabrication. Through keynote talks, lightning presentations, and a moderated panel, it will showcase both what differentiable physics can already enable and the technical barriers that still limit broader adoption, including efficiency, robustness, contact-rich dynamics, and the gap between application-facing representations and simulation-ready assets.
Attendees can expect an engaging forum that connects developers with researchers and practitioners seeking to use simulation as an optimization and learning primitive, not just a forward model. A key merit of the workshop is its ability to bridge communities that do not often meet in one venue, helping participants identify practical opportunities, clarify open problems, and shape future directions for differentiable simulation in graphics and AI.
The workshop is organized by researchers from Carnegie Mellon University, UCLA, the University of Utah, and Meta Reality Labs, whose work spans physical simulation, graphics systems, robotics, and AI. Together with invited speakers from academia and industry, they bring perspectives from foundational simulation research to large-scale production systems and real-world applications.
The workshop brings together foundational methods, practical systems, and emerging applications in differentiable simulation across graphics, robotics, 3D vision, design, and fabrication. Through keynote talks, lightning presentations, and a moderated panel, it will showcase both what differentiable physics can already enable and the technical barriers that still limit broader adoption, including efficiency, robustness, contact-rich dynamics, and the gap between application-facing representations and simulation-ready assets.
Attendees can expect an engaging forum that connects developers with researchers and practitioners seeking to use simulation as an optimization and learning primitive, not just a forward model. A key merit of the workshop is its ability to bridge communities that do not often meet in one venue, helping participants identify practical opportunities, clarify open problems, and shape future directions for differentiable simulation in graphics and AI.
Organizers
Speakers
Event Type
Technical Workshop
TimeMonday, 20 July 20262:00pm - 5:15pm PDT
LocationRoom 406 AB
Full Conference Supporter
Full Conference
