Research Analyzer
← Back ICRA 2026

GAPG: Geometry Aware Push-Grasping Synergy for Goal-Oriented Manipulation in Clutter

Lijingze Xiao, jinhong Du, YANG CONG, Supeng Diao, Yu Ren

PDF

AI summary

Key figure (auto-extracted from paper)
Leveraging 3D point cloud geometry to jointly evaluate grasp feasibility and push effectiveness enables reliable, collision-free target grasping in cluttered scenes without fine-tuning.
Point cloud processing Push-grasp synergy Cluttered scene manipulation Geometric grasp evaluation Sim-to-real transfer

Problem

Single-step grasping fails in cluttered scenes due to occlusion and limited space, while existing push-grasp methods neglect 3D geometric information, resulting in unstable grasps and inefficient pushing.

Approach

GAPG uses 3D point clouds to jointly evaluate grasp feasibility and push effectiveness via two PointNet++-based modules that analyze spatial geometry and gripper-object alignment to guide safe, efficient manipulation.

Key results

  • Accurate grasp feasibility prediction via gripper-target geometric alignment
  • Effective push action selection that reliably creates graspable space
  • Strong sim-to-real transfer and generalization to unseen objects without fine-tuning
  • Superior task completion and evaluation accuracy over baseline methods in cluttered scenes

Why it matters

This work advances goal-oriented robotic manipulation by providing a geometry-aware framework that enables reliable, collision-free grasping in unstructured environments, benefiting warehouse automation and home service robotics.

Abstract

Grasping target objects is a fundamental skill for robotic manipulation, but in cluttered environments with stacked or occluded objects, a single-step grasp is often insuf- ficient. To address this, previous work has introduced pushing as an auxiliary action to create graspable space. However, these methods often struggle with both stability and efficiency because they neglect the scene’s geometric information, which is essential for evaluating grasp robustness and ensuring that pushing actions are safe and effective. To this end, we propose a geometry-aware push–grasp synergy framework that leverages point cloud data to integrate grasp and push evaluation. Specifically, the grasp evaluation module analyzes the geometric relationship between the gripper’s point cloud and the points enclosed within its closing region to determine grasp feasibility and stability. Guided by this, the push evaluation module predicts how pushing actions influence future graspable space, enabling the robot to select actions that reliably transform non-graspable states into graspable ones. By jointly reasoning about geometry in both grasping and pushing, our framework achieves safer, more efficient, and more reliable manipulation in cluttered settings. Our method is extensively tested in simulation and real-world environments in various scenarios. Experimental results demonstrate that our model generalizes well to real- world scenes and unseen objects. The code and video are available at https://github.com/xiaolijz/GAPG.

Index terms

Grasping Deep Learning for Visual Perception Perception for Grasping and Manipulation

Related papers