Ever tried to get a robot to fully map an object that changes shape the second it touches it? 🤔 Autonomous exploration of unknown soft objects is notoriously difficult because they deform unpredictably, keeping much of their surface hidden or self-occluded. To crack this problem, our new paper in the The International Journal of Robotics Research (IJRR) introduces a novel framework integrating topology-aware 3D reconstruction with a topology-guided motion planner.
Under the hood:
🧠 Real-time Topology Tracking: Combines Cylinder Čech Complexes with point clouds to instantly track drastic topology changes and fix non-manifold geometries.
🦾 Smart Active Manipulation: Uses a Null Space Criterion to optimize grasp points, actively executing targeted “turning over” and “stretching” actions to expose hidden surfaces.
🫀 Real-world Validation: Achieved full homeomorphic reconstruction on complex items, including actual human cadaveric liver and colon tissues, using the da Vinci Research Kit (dVRK).
Huge thanks to co-authors Dominic Jones, Gerardo Loza, Shoudong Huang, Pietro Valdastri and the team at STORM lab!
🔗 Read the full paper here: https://journals.sagepub.com/doi/10.1177/02783649251415415