Award-Winning NVIDIA Robotics Research Takes Center Stage at ICRA

At the International Conference on Robotics and Automation, NVIDIA demonstrates the power of robotics simulation for real-world use cases.
by Jason Black

More than a dozen NVIDIA robotics research papers will take center stage at this year’s IEEE International Conference on Robotics and Automation (ICRA), taking place May 29-June 2, in London.

One, titled “Geometric Fabrics: Generalizing Classical Mechanics to Capture the Physics of Behavior,” will be recognized with an IEEE Best Paper award.

Many of this year’s submissions demonstrate robotics capabilities trained in simulation that were then shown to work in the real world.

Additional papers from members of the NVIDIA Robotics Research Lab to be presented at ICRA include:

  • “DeXtreme: Transferring Agile In-Hand Manipulation From Simulation to Reality” (Read paper, watch video)
  • “MVTrans: Multi-View Perception of Transparent Objects” (Read paper, watch video)
  • “Self-Supervised Learning of Action Affordances as Interaction Modes” (Read paper, watch video)
  • “nerf2nerf: Pairwise Registration of Neural Radiance Fields” (Read paper, watch video)
  • “DexGrasp-1M: Dexterous Multi-Finger Grasp Generation Through Differentiable Simulation” (Read paper, watch video)

  • “ProgPrompt: Generating Situated Robot Task Plans Using Large Language Models” (Read paper, watch video)
  • “DefGraspNets: Grasp Planning on 3D Fields With Graph Neural Nets” (Read paper, watch video)
  • “CuRobo: Parellelized Collision-Free Robot Motion Generation” (Read paper)
  • “RGB-Only Reconstruction of Tabletop Scenes for Collision-Free Manipulator Control” (Read paper, watch video)
  • “FewSOL: A Dataset for Few-Shot Object Learning in Robotic Environments” (Read paper, watch video)

  • “Planning for Multi-Object Manipulation With Graph Neural Network Relational Classifiers” (Read paper)
  • “CabiNet: Scaling Neural Collision Detection for Object Rearrangement With Procedural Scene Generation” (Read paper, watch video)
  • “Global and Reactive Motion Generation With Geometric Fabric Command Sequences” (Read paper, watch video)

Many of the papers above include GitHub code links for those who want to explore.

Visit NVIDIA partners at ICRA — including, Cogniteam, Connect Tech, Mathworks and Silicon Highway. They’ll be showcasing real-world robotics research demos, powered by the NVIDIA Jetson Orin platform for edge AI and robotics as well as the NVIDIA Isaac robotics simulation platform.

In addition, check out this technical blog on how robotics software company Cogniteam set up a multi-robot simulation scenario using its Nimbus platform and NVIDIA Isaac Sim. The setup offered capabilities such as local simulation tasks, remote access to simulation machines, global monitoring of simulation data and more — all accessible through a web browser.

Learn more about the NVIDIA Robotics Research Lab, a Seattle-based center of excellence focused on robot manipulation, perception and physics-based simulation. It’s part of NVIDIA Research, which comprises more than 300 leading researchers around the globe, focused on topics spanning AI, computer graphics, computer vision and self-driving cars.