DG16M

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.

Video Explanation

Dataset Generation Pipeline

Overview of the proposed method: (a) We start by sampling a large number of antipodal grasps on the object mesh, generate all possible grasp pairs, and apply distance-based pruning. These pairs are evaluated using an Optimizer-based Force Closure Evaluator, which checks if a valid set of contact forces can keep the object in equilibrium under an external wrench. (b) The optimizer solves the force closure problem by minimizing the contact forces constrained to gripper force limits and friction cone constraints. Valid grasp pairs are identified by thresholding the loss values, ensuring that only stable grasps are retained while unstable ones are discarded. (c) We select a subset of objects and evaluate their grasps in simulation to construct the benchmark dataset. This evaluation provides stability ground truth values for each dual-arm grasp, offering an objective assessment that is independent of predefined grasp quality metrics or assumptions.

Qualitative Results (coming soon)

BibTeX

@misc{dg16m2025,
    title={DG16M: A Large-Scale Dataset for Dual-Arm Grasping with Force-Optimized Grasps}
    author={Md Faizal Karim and Mohammed Saad Hashmi and Shreya Bollimuntha and Mahesh Reddy Tapeti and Gaurav Singh and Nagamanikandan Govindan and K Madhava Krishna},
    year={2025},
    journal={}
    
}