DG16M: A Large-Scale Dataset for Dual-Arm Grasping with Force-Optimized Grasps

2025-03-11

AI Summary

Dual-arm grasping is hindered by the scarcity of large-scale, high-confidence annotated datasets—particularly those grounded in rigorous force-closure stability criteria. To address this, we introduce the first large-scale force-optimized dataset for dual-arm grasping (16 million samples), propose an enhanced force-closure evaluation paradigm, and release a standardized simulation benchmark comprising 300 diverse objects and 30,000 physically validated grasps. Our framework establishes the first end-to-end integration of analytical force-closure analysis, high-fidelity physics simulation (PyBullet/Isaac Gym), and deep learning via a Dual-Arm Grasp Classifier. Evaluated under cross-object generalization, our trained classifier achieves a 15% absolute improvement in grasp success rate over state-of-the-art methods. This work significantly advances the reliability, interpretability, and quantitative assessment of dual-arm robotic manipulation.

Technology Category

• Intelligent Robots: Manipulation
• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Security and Privacy: Large-scale security measurements
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Dual-arm robotic grasping is crucial for handling large objects that require stable and coordinated manipulation. While single-arm grasping has been extensively studied, datasets tailored for dual-arm settings remain scarce. We introduce a large-scale dataset of 16 million dual-arm grasps, evaluated under improved force-closure constraints. Additionally, we develop a benchmark dataset containing 300 objects with approximately 30,000 grasps, evaluated in a physics simulation environment, providing a better grasp quality assessment for dual-arm grasp synthesis methods. Finally, we demonstrate the effectiveness of our dataset by training a Dual-Arm Grasp Classifier network that outperforms the state-of-the-art methods by 15%, achieving higher grasp success rates and improved generalization across objects.

Problem

• Lack of datasets for dual-arm robotic grasping scenarios.
• Need for improved force-closure constraints in grasp evaluation.
• Requirement for better grasp quality assessment in dual-arm manipulation.

Innovation

• Large-scale dataset with 16 million dual-arm grasps
• Improved force-closure constraints for evaluation
• Dual-Arm Grasp Classifier outperforms state-of-the-art by 15%