analiza aiqbench
MultiGraspNet advances robotic automation by introducing a unified 3D deep learning framework that simultaneously predicts grasp poses for both parallel and vacuum grippers. This multitask architecture allows a single robot to handle diverse end-effectors efficiently, leveraging shared early-stage features while maintaining specific refiners for each modality. The model achieves competitive performance on standard benchmarks with reduced computational cost, demonstrating significant potential for industrial applications requiring versatile manipulation in cluttered environments.
MultiGraspNet: A Multitask 3D Vision Model for Multi-gripper Robotic Grasping
arXiv:2602.06504v2 Announce Type: replace Abstract: Vision-based models for robotic grasping automate critical, repetitive, and draining industrial tasks. Existing approaches are typically limited in two ways: they either target a single gripper and are potentially applied on costly dual-arm setups, or rely on custom hybrid grippers that require ad-hoc learning procedures with logic that cannot be transferred across tasks, restricting their general applicability. In this work, we present MultiGraspNet, a novel multitask 3D deep learning method that predicts feasible poses simultaneously for parallel and vacuum grippers within a unified framework, enabling a single robot to handle multiple end effectors. The model is trained on the richly annotated GraspNet-1Billion and SuctionNet-1Billion datasets, which have been aligned for the purpose, and generates graspability masks quantifying the suitability of each scene point for successful grasps. By sharing early-stage features while maintaining gripper-specific refiners, MultiGraspNet effectively leverages complementary information across grasping modalities. This design preserves a compact architectural footprint of only 15.75M parameters and enables fast inference on a single GPU, enhancing adaptability and efficiency in cluttered scenes. We characterize MultiGraspnet's performance with an extensive experimental analysis, demonstrating its competitiveness with single-task models on relevant benchmarks while reducing computational cost. Moreover, real-world experiments on a single-arm multi-gripper robotic setup show that our approach outperforms normalization-based multi-gripper approaches. Project page: https://vandal-lab.github.io/multigraspnet-project
Abstract:Vision-based models for robotic grasping automate critical, repetitive, and draining industrial tasks. Existing approaches are typically limited in two ways: they either target a single gripper and are potentially applied on costly dual-arm setups, or rely on custom hybrid grippers that require ad-hoc learning procedures with logic that cannot be transferred across tasks, restricting their general applicability. In this work, we present MultiGraspNet, a novel multitask 3D deep learning method that predicts feasible poses simultaneously for parallel and vacuum grippers within a unified framework, enabling a single robot to handle multiple end effectors. The model is trained on the richly annotated GraspNet-1Billion and SuctionNet-1Billion datasets, which have been aligned for the purpose, and generates graspability masks quantifying the suitability of each scene point for successful grasps. By sharing early-stage features while maintaining gripper-specific refiners, MultiGraspNet effectively leverages complementary information across grasping modalities. This design preserves a compact architectural footprint of only 15.75M parameters and enables fast inference on a single GPU, enhancing adaptability and efficiency in cluttered scenes. We characterize MultiGraspnet's performance with an extensive experimental analysis, demonstrating its competitiveness with single-task models on relevant benchmarks while reducing computational cost. Moreover, real-world experiments on a single-arm multi-gripper robotic setup show that our approach outperforms normalization-based multi-gripper approaches. Project page: this https URL
Accepted for publication in IEEE Robotics and Automation Letters (2026). 8 pages, 5 figures
Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
arXiv:2602.06504 [cs.RO]
(or arXiv:2602.06504v2 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2602.06504 Focus to learn more arXiv-issued DOI via DataCite
https://doi.org/10.1109/LRA.2026.3719235 Focus to learn more DOI(s) linking to related resources
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 333 cuvinte
Subiecte conexe
The Machines that Make the Machines
NVIDIA · 7 oct. 2026
How to Evaluate General-Purpose Robot Policies for Real-World Deployment
NVIDIA · 12 iul. 2026
Building a Foundation Stack for General-Purpose Robots
IEEE Spectrum · 13 iul. 2026
Temporal Visuo-Tactile Learning for Dexterous Grasp Stability
arXiv · 8 oct. 2026