Robots need to be able to grab and manipulate a wide variety of items. Researchers are always working for ways to improve them. GraspGen is a 6D pose generation and planning approach that allows for fast object segmentation, pose estimation and grasp planning. It works on desktop PCs and Jetson. For this project, a PiPER Robotic arm was used.
As the researchers explain:
GraspGen – consists of a Diffusion-Transformer 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.
You can read the paper here.
[HT: Murali, Adithyavairavan and Sundaralingam, Balakumar and Chao, Yu-Wei and Yamada, Jun and Yuan, Wentao and Carlson, Mark and Ramos, Fabio and Birchfield, Stan and Fox, Dieter and Eppner, Clemens]
