SOFA Framework

🏆✨ We are thrilled to share that our poster has received the Best Poster Award at the IROS 2026 Workshop on Mechanically Intelligent Soft Robotics! 🤖🎉 🔗 Workshop website: https://lnkd.in/emNpVpSi 🧩 Our awarded work, “Mechanical Intelligence in Soft Robotic Lattices,” explores how geometry and architecture can embed actuation, pneumatic routing, sensing, and control directly into soft robotic structures. Through programmable lattice designs and embedded pneumatic unit cells, we show how soft robots can achieve: 🔹 Programmable shape changes and morphologies 🔹 Integrated pneumatic actuation and sensing 🔹 Recognition of global deformation versus localized contact 🔹 Mechanically informed robotic behaviors, including crawling and tactile interaction ⚙️This work moves toward a future in which intelligence is not only computed electronically, but also designed into the robot’s physical body and material architecture. 👏 Congratulations to all the authors: Altair Coutinho, Trevor Exley, Petr Trunin, Anderson Nardin, and Lucia Beccai. 🎓🔬 🙏 A sincere thank you to the workshop organizers and all participants for the inspiring discussions. We are honored by this recognition! #IROS2026 #SoftRobotics #MechanicalIntelligence #RoboticLattices #EmbodiedIntelligence #PneumaticRobotics #BioRobotics #RoboticsResearch

🤖 What if surgical simulation could generate not just motion data—but physically grounded interaction data? I’m excited to share that I’ll be presenting our work, “Semi-Automated Motion Primitives for Force-aware Rigid and Deformable Manipulation in Simulated Robotic Surgery,” at the Evolving Landscape for Surgical Robotics (ELSR) Workshop at IROS 2026. 📍 ELSR Workshop @ IROS 2026 🏛️ David L. Lawrence Convention Center, Pittsburgh, PA 🚪Room 410 🗓️ Thursday, October 1st, 2026 🕘 8:30 AM – 5:30 PM 🔬 The idea: Learning-based surgical autonomy requires large amounts of interaction data, but synthetic data is only useful when the underlying trajectories and interactions are physically meaningful and repeatable. We developed a semi-automated motion primitive framework for the da Vinci Research Kit (dVRK) in the SOFA multiphysics engine, using: • Reach, Path Follow, Jaw Open/Close, Homing • Composable Approach, Grasp, Transport, Release primitives • Deterministic proximity-based collision avoidance • A unified Grasp-Lock mechanism for rigid and deformable objects ⚙️ The key aspect: Force-aware manipulation directly from the physics layer. Rather than adding a simulated force sensor, we recover constraint-derived contact forces from Lagrange multipliers in the SOFA constraint solver. This allows us to track grasp onset, contact events, loading, and force signatures during manipulation of both rigid and FEM-based deformable bodies. For rigid objects, Grasp-Lock maintains a fixed SE(3) grasp relationship. For deformable objects, FEM nodes within an instrument-derived capture region are constrained to the grasp frame—allowing the same manipulation primitives to operate across rigid and deformable objects. Please visit our project website for more information: https://lnkd.in/gBJxmiBZ 🚀 The Broader Goal This work explores a middle ground between classical robotic control, physics-based simulation, and learning-based surgical autonomy—using repeatable, physics-grounded motion primitives to generate and evaluate manipulation trajectories before learned policies are exposed to greater task variability. Thanks to my co-authors Dr. Noah Shaikh, Dr. Ahmed H. Qureshi, Dr. Dimitrios Stefanidis, and Dr. Sohel Anwar for their immense guidance and support. Looking forward to sharing the work and discussing the evolving role of physics-based simulation, synthetic data, and force-aware manipulation in surgical robotics. #IROS2026 #SOFA #dVRK #IntuitiveSurgical #ContactMechanics #MotionPrimitives #Multiphysics #LagrangeMultipliers #FiniteElementModeling #Deformations #SyntheticData #SoftTissueGrasping #SurgicalSimulation #RoboticSurgery #SurgicalRobotics #RoboticLaparoscopy #PurdueEngineering

🫀What does it take for a surgical digital twin to be biomechanically trustworthy? I’m excited to present our work today at the SurgTwin Workshop at IROS 2026 — “From Generative AI to Digital Twins: Perception, Simulation, and Learning for Surgical Robotics.” Our work, “FEM Paradigms for Heterogeneity Modeling in Multi-Material Tissue-Tumor Surgical Twins,” explores a question that I think is fundamental to physics-based surgical digital twins: Can a simulated tumor be mechanically meaningful enough that a robotic system can actually interrogate it through palpation? 🔍 The problem: Robotic surgery gives surgeons great vision but removes touch. Tumors below the organ surface are invisible to the camera. To rehearse force-aware palpation before surgery, we need digital twins whose physics are trustworthy. ⚙️ The challenge Modeling a stiff tumor inside soft tissue usually means two separate meshes held together by friction or constraints. Under gravity and instrument contact, they tear, interpenetrate or separate without proper physics tuning per model. 🧪 What we did: In SOFA (Multiphysics engine), we built a CT-derived liver-tumor twin palpated by dVRK instruments and compared three FEM formulations: ▪️ Uncoupled multibody: 100% interface failure at full gravity ▪️ Constraint-coupled multibody: 73% failure. ▪️ Single-mesh Multi-material: 0% failure, with continuity by construction. Palpation detected a 1.24 to 1.30 apparent stiffness contrast, consistent with physical attenuation over a buried tumor. 📊 Why it matters: New Liver-tumor digital twin environments can be generated through an automated pipeline. By sweeping tumor size, depth and stiffness, we get synthetic force-deformation and dVRK kinematic data , ready for training RL palpation policies and tumor boundary estimation for pre-surgical assessment. I would like to thank my co-authors: Dr. Noah Shaikh, Dr. Dimitrios Stefanidis, Dr. Ahmed H. Qureshi and Dr. Sohel Anwar, for their immense support and guidance. Please look at our project page for more information: https://lnkd.in/gPsrYXTG You can find me at: SurgTwin Workshop, IROS 2026 📍 Room: 336 🗓️ Sunday, September 27, 2026 🕘 1:30 pm to 5:30 pm 🏛️ David L. Lawrence Convention Center, Pittsburgh, PA. If you are at IROS, please stop by! I would love to talk about surgical digital twins, soft tissue simulation, and sim-to-real transfer for surgical robotics and deformable body manipulation. 🤖 #IROS2026 #DigitalTwin #SOFA #dVRK #IntuitiveSurgical #TactilePerception #RoboticPalpation #SyntheticData #SimToReal #SurgicalSimulation #SurgicalRobotics #MedicalRobotics #TumorLocalization #FiniteElementModeling #RoboticSurgery #PhDResearch #PurdueEngineering

2ieme jour sur le salon SIDO Lyon - IoT, IA, Robotique & XR Démonstration de notre robot souple associé à notre jumeau numérique Compliance Robotics est impatient de vous rencontrer !

🤖 We are excited to announce the release of PiecewiseElasticity, a free and open-source plugin developed by Anderson Nardin. 🧩 PiecewiseElasticity extends SOFA Framework with TetrahedronFEMForceFieldPWE, a tetrahedral FEM force field that adapts Young’s modulus according to measured strain. ⚙️ Users can define strain thresholds and corresponding stiffness values, with optional smoothing between intervals. This makes it possible to model materials whose mechanical response changes across deformation regimes—for example, experimentally characterized or homogenized architected materials with different stiffness in tension and compression. 🛠️ The repository includes installation instructions, tension/compression simulation examples, and comparison plots against the standard constant-modulus FEM component. 🔗 GitHub: https://lnkd.in/e8T9BhaB 💡 We welcome feedback, ideas, and contributions from the soft robotics and simulation communities. #SoftRobotics #SOFAFramework #FiniteElementMethod #Biomechanics #Simulation #OpenSource #ComputationalModeling #SoftBioRobotics #Robotics

Still experimenting with catheter navigation, this time in SOFA Framework with the BeamAdapter plugin. The clip shows a generic three-section catheter: a stiff proximal shaft, a softer deflectable segment, and a very soft tip. It is simulated navigating, then navigation while rotating, through a tortuous path based on the fixture in ASTM F2394, on a plain CPU with friction contact. The interesting part is the handoff. BeamAdapter wants a Young's modulus, a Poisson ratio and radii for each section. A real catheter does not have those numbers. It has bending, torsional and axial stiffness that come from the braid, the liner and the jacket. To be clear: navigation is not shipping in VirtuCath any time soon. But if you are building endoluminal robotics, navigation training, RL environments or patient-specific planning, your simulation needs a catheter that responds like the real device. A single stiffness value for the whole shaft cannot represent a braided shaft with a soft tip. VirtuCath exports the calculated section properties instead: - Per-section mechanics: EI, GJ, EA and mass per unit length along the full arclength. Derived from braid PPI, wire geometry, coil, liner and jacket durometer. One JSON file, documented schema. - Reads into solvers you already run: SOFA with BeamAdapter, NVIDIA Isaac for Healthcare, or MuJoCo. - Scales to training runs: a scriptable CLI turns manufacturing tolerances on PPI, material properties, section lengths or wall thickness into fleets of device variants. A tolerance on the braid moves EI and GJ together the way a real production lot drifts, which adding noise to each stiffness independently never does. If you are working on this, I would love to chat. #MedTech #MedicalDevices #SurgicalRobotics #Simulation #CatheterDesign