GraspGen: A Diffusion-based Framework for 6-DOF Grasping with On-Generator Training
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908454726139904 |
|---|---|
| author | Murali, Adithyavairavan Sundaralingam, Balakumar Chao, Yu-Wei Yuan, Wentao Yamada, Jun Carlson, Mark Ramos, Fabio Birchfield, Stan Fox, Dieter Eppner, Clemens |
| author_facet | Murali, Adithyavairavan Sundaralingam, Balakumar Chao, Yu-Wei Yuan, Wentao Yamada, Jun Carlson, Mark Ramos, Fabio Birchfield, Stan Fox, Dieter Eppner, Clemens |
| contents | Grasping is a fundamental robot skill, yet despite significant research advancements, learning-based 6-DOF grasping approaches are still not turnkey and struggle to generalize across different embodiments and in-the-wild settings. We build upon the recent success on modeling the object-centric grasp generation process as an iterative diffusion process. Our proposed framework, GraspGen, consists of a DiffusionTransformer architecture that enhances grasp generation, paired with an efficient discriminator to score and filter sampled grasps. We introduce a novel and performant on-generator training recipe for the discriminator. To scale GraspGen to both objects and grippers, we release a new simulated dataset consisting of over 53 million grasps. We demonstrate that GraspGen outperforms prior methods in simulations with singulated objects across different grippers, achieves state-of-the-art performance on the FetchBench grasping benchmark, and performs well on a real robot with noisy visual observations. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2507_13097 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | GraspGen: A Diffusion-based Framework for 6-DOF Grasping with On-Generator Training Murali, Adithyavairavan Sundaralingam, Balakumar Chao, Yu-Wei Yuan, Wentao Yamada, Jun Carlson, Mark Ramos, Fabio Birchfield, Stan Fox, Dieter Eppner, Clemens Robotics Artificial Intelligence Grasping is a fundamental robot skill, yet despite significant research advancements, learning-based 6-DOF grasping approaches are still not turnkey and struggle to generalize across different embodiments and in-the-wild settings. We build upon the recent success on modeling the object-centric grasp generation process as an iterative diffusion process. Our proposed framework, GraspGen, consists of a DiffusionTransformer architecture that enhances grasp generation, paired with an efficient discriminator to score and filter sampled grasps. We introduce a novel and performant on-generator training recipe for the discriminator. To scale GraspGen to both objects and grippers, we release a new simulated dataset consisting of over 53 million grasps. We demonstrate that GraspGen outperforms prior methods in simulations with singulated objects across different grippers, achieves state-of-the-art performance on the FetchBench grasping benchmark, and performs well on a real robot with noisy visual observations. |
| title | GraspGen: A Diffusion-based Framework for 6-DOF Grasping with On-Generator Training |
| topic | Robotics Artificial Intelligence |
| url | https://arxiv.org/abs/2507.13097 |