Dexterous robot hands learn fast grasping in cluttered spaces
T(R,O) Grasp++: Cross-Embodiment Dexterous Grasping in Cluttered Environments at Scale
Robotics
Summary
Grasping objects with robot hands is hard, especially when things are crowded together. The authors created a huge dataset of simulated hand grasps and objects to help teach robots how to do this. They also made a new approach that helps different robot hands learn to pick up items quickly and accurately. Their method works well across many types of robot hands and is much faster than older ones.
What this means in practice
- •For robotics engineers: Enable robots with different hands to quickly learn to pick up objects in crowded, real-world settings.$Commercial implications: Provides scalable grasp synthesis methods for manufacturing and logistics robots, improving object handling capabilities and speed.
- •For automation system builders: Integrate a fast and efficient grasp planning pipeline that supports multiple robotic hand designs in cluttered environments.
Authors
Xin Fei, Zhixuan Xu, Huaicong Fang, Fangjun Chen, Zhaoyu Mu, Rui Yang, Zhexi Luo, Lin Shao
Abstract
Dexterous grasping in cluttered scenes remains a central challenge in robotic manipulation, as it requires high-dimensional hand coordination under complex object-environment constraints. To address this problem, we introduce one of the largest simulation benchmarks to date, comprising around 0.5B grasp annotations across eight dexterous hands and approximately 50K objects. Furthermore, we propose T(R,O) Grasp++, a scalable diffusion-based framework for cross-embodiment dexterous grasp synthesis in cluttered environments. Extensive experiments demonstrate that our model significantly outperforms existing baselines, achieving an average success rate of 88% across eight dexterous embodiments. Meanwhile, our pipeline achieves only 0.3s inference latency while improving inference sample efficiency by over 10 times.