SIGGRAPH Asia 2026 · Technical Workshop

Beyond Rigid Bodies

Next-Generation Differentiable Simulation for Physical AI and Embodied Robotics — faithful, fast, differentiable, and deployable physics beyond rigid-body dynamics, at the intersection of computer graphics and robotics.

  • When Dec 1–4, 2026 (exact day TBD)
  • Where Kuala Lumpur, Malaysia
  • Format Half day · ~3.5 hrs
Four robot-manipulation panels: (1) a hand pressing soft tissue, (2) a dexterous hand reorienting a soft object in-hand, (3) cloth reconstructed from a camera into a simulated digital twin, and (4) a deformable cable reshaped toward a goal via predicted rollouts.
Beyond Rigid Bodies across four contact-rich robot-manipulation settings, each hinging on physics that rigid-body simulators cannot represent.

Overview

Simulation, beyond rigid bodies.

The intersection of computer graphics and robotics has seen immense progress, yet most scalable Embodied AI still relies on rigid-body dynamics. As robots increasingly interact with soft tissue, granular media, deformable objects, and high-precision frictional contact, there is a critical need for next-generation physical simulators that are simultaneously faithful, fast, differentiable, and deployable.

Beyond Rigid Bodies (BRB) convenes researchers across computational mechanics, differentiable simulation, and robot learning. The program pairs invited keynote deep dives with extended, structured discussion and a closing panel, organized around four themes.

Four pillars

The technical agenda

i FEM / IPC contact

High-Fidelity Non-Rigid Simulation

FEM, MPM, and IPC solvers; frictional contact; near-incompressible and hyperelastic materials; granular and elastoplastic media; cloth and thin shells; GPU-accelerated, real-time solvers.

ii θ contact

Differentiable Simulation for Policy Learning

Analytic gradients through contact for system identification, inverse design, computational design and fabrication, and robot manipulation and control.

iii Real → Sim

Real2Sim2Real & Digital Twins

Neural radiance fields and Gaussian splatting; scene and dynamics reconstruction; Sim-to-Real transfer; calibrated, deployable closed-loop pipelines.

iv predict + plan

Simulation as a World Model

Neural and generative world models; learned physics surrogates; neural PDEs/ODEs and reduced-order modeling; multi-modal predictive models for embodied agents.

Keynote speakers

Four invited deep dives

Taku Komura
Keynote

Taku Komura

Professor · The University of Hong Kong

character animation · physics-based simulation · contact solvers

Leads research in character animation, physics-based simulation, and interactive techniques. Recent work on whole-body manipulation, contact-rich scene interaction, and GPU-accelerated contact solvers sits at the boundary of animation and robot learning.

Tao Du
Keynote

Tao Du

Assistant Professor · Tsinghua University (IIIS)

differentiable simulation · soft robots · inverse design

Develops computational design frameworks and differentiable simulation tools for soft robots, actuated mechanisms, and fabrication-aware optimization, discovering robot morphologies and control policies directly from simulation.

Yunzhu Li
Keynote

Yunzhu Li

Assistant Professor · Columbia University (RoboPIL)

world models · deformable manipulation · Real2Sim

Works on structured world models and physics-inspired predictive models for Embodied AI, with emphasis on multi-modal perception, manipulation of deformable objects, and bridging learned simulators with real-world deployment.

Srinath Sridhar
Keynote

Srinath Sridhar

Assistant Professor · Brown University (IVL)

3D/4D vision · hand–object · dynamic neural fields

Builds foundational methods for 3D/4D understanding of objects, humans in motion, and hand–object interaction. Work on dynamic neural fields, articulated-object understanding, and visuo-haptic pose tracking connects to Real2Sim and digital twins.

Schedule

Half day · ~3.5 hours

Keynotes run ~35 minutes including discussion. Exact date, room, and start time will be confirmed by SIGGRAPH Asia 2026.

  1. 0:00–0:15

    Opening Remarks

    Organizers

  2. 0:15–0:50

    Keynote 1

    Taku Komura

  3. 0:50–1:25

    Keynote 2

    Tao Du

  4. 1:25–1:40

    Coffee Break

    All attendees

  5. 1:40–2:15

    Keynote 3

    Srinath Sridhar

  6. 2:15–2:50

    Keynote 4

    Yunzhu Li

  7. 2:50–3:30

    Panel & Open Discussion

    All speakers + organizers — open problems, benchmarks, practical barriers, and collaboration

Organizers

The organizing team

Connecting graphics simulation, computational mechanics, 3D vision, robotics, and physical AI across academia and industry.

Shih-Yu Lai

Shih-Yu Lai

Researcher · NTU & MoonShine Animation Studio

Ziqiu Zeng

Ziqiu Zeng

Research Fellow · National University of Singapore

Fan Shi

Fan Shi

Assistant Professor · National University of Singapore

Peter Yichen Chen

Peter Yichen Chen

Assistant Professor · University of British Columbia

Tiantian Liu

Tiantian Liu

Independent Researcher

Yu-Lun Liu

Yu-Lun Liu

Associate Professor · National Yang Ming Chiao Tung University

Bing-Yu Chen

Bing-Yu Chen

Distinguished Professor · National Taiwan University

Institutions & partners

Organizing institutions

National Taiwan University National University of Singapore University of British Columbia National Yang Ming Chiao Tung University The University of Hong Kong Tsinghua University Columbia University Brown University

Supported by

MoonShine Animation Studio Shanda NVIDIA Prana Labs

Bring physics into the loop.

Join us at SIGGRAPH Asia 2026 in Kuala Lumpur (Dec 1–4, 2026) for four keynote deep dives and an extended panel on faithful, fast, differentiable, and deployable simulation.

Contact: akinesia112@gmail.com