PartialBiGrasp: Inferring Hidden Local Geometry for Bimanual Grasping from Partial Views

Noosaga Placements

• Robot Manipulationsubfield95%
  The paper focuses on dual-arm grasp generation and refinement, which is a core topic in robot manipulation.
  Dual-arm robotic grasping is essential for manipulating large, heavy, and geometrically complex objectswe propose PartialBiGrasp, a dual-arm grasp generation framework
• Analytical Grasping and Fixturingframework90%
  The paper uses analytical force-closure constraints to evaluate and generate grasp pairs, which is a key concept in analytical grasping.
  generate force-closure compliant grasp pairsevaluate our approach using analytical force-closure metrics
• Data-Driven and Learning-Based Manipulationframework90%
  The method is a learning-based approach for grasp generation, fitting the data-driven manipulation framework.
  Our model learns geometric features implicitly through convolutional occupancy networkslearning-based bimanual grasp generation
• Roboticssubfield80%
  The work is a robotics application that uses learning-based methods, fitting under the broader AI robotics umbrella.
  Our model learns geometric features implicitly through convolutional occupancy networksevaluate our approach using analytical force-closure metrics, large-scale simulation experiments, and real-world robot evaluations
• Learning-Based Roboticsframework80%
  The paper uses a learning-based approach for robotic grasping, which is a key aspect of learning-based robotics.
  Our model learns geometric features implicitly through convolutional occupancy networkslearning-based bimanual grasp generation
• Robot Learningsubfield70%
  The method involves training neural networks for grasp generation and refinement, which is a form of robot learning.
  Our model learns geometric features implicitly through convolutional occupancy networksTraining is performed in three stages
• Convolutional Neural Networksframework80%
  The method uses convolutional occupancy networks, which are a type of convolutional neural network.
  learns geometric features implicitly through convolutional occupancy networks
• Computer Visionsubfield60%
  The method processes partial point clouds and uses occupancy networks, which are common in computer vision for 3D understanding.
  operates directly on partial point cloud observationslearns geometric features implicitly through convolutional occupancy networks
• Geometric and Physical Reconstructionframework70%
  The paper infers hidden geometry from partial observations, which is a form of geometric reconstruction.
  reasoning about hidden local geometryinferring hidden local geometry

Abstract

Dual-arm robotic grasping is essential for manipulating large, heavy, and geometrically complex objects that cannot be reliably handled using a single manipulator. These large objects often contain only sparse graspable regions determined by local geometric properties such as thickness, edge structure, and gripper clearance. Prior bimanual grasping methods assume access to a full point cloud of the object which inherently contains this geometric information, but may not be accessible in real scenarios. This work proposes PartialBiGrasp, a dual-arm grasp generation framework that operates directly on partial point cloud observations. Our model learns geometric features implicitly through convolutional occupancy networks, enabling local reasoning about graspability, collision-free contact regions, and object thickness. We leverage this understanding to generate force-closure compliant grasp pairs, which are further refined using a sampling-based optimization to correct for ambiguity caused by incomplete geometry. We evaluate our approach using analytical force-closure metrics, large-scale simulation experiments, and real-world robot evaluations on noisy partial point clouds of novel objects, demonstrating robust and physically stable dual-arm grasp generation.

Paper Context