MultiGraspNet

MultiGraspNet — A Multitask 3D Vision Model for Multi-gripper Robotic Grasping

We present MultiGraspNet, a novel multi-task 3D deep learning method that predicts feasible poses simultaneously for parallel and vacuum grippers within a unified framework, enabling a single robot to handle both end-effectors. By sharing early-stage features while maintaining gripper-specific refiners, the model effectivelly leverages complementary information across grasping modalities with a compact footprint of only 15.75M parameters and fast inference on a single GPU. Real-world experiments on a single-arm multi-gripper robotic setup demonstrate competitive performance against single-task baselines while reducing computational cost.
Paper Preprint Code (coming soon)

Project Video

Overview of the MultiGraspNet project, including the full pipeline and a demo in real-world.

Visualizations

Interactive 3D point cloud visualization. Click and drag to rotate and scroll to zoom.

Get in Touch

For questions about the paper, method, or dataset alignment, contact the corresponding author directly.

stephany.ortuno@polito.it

Acknowledgements

This work was supported by Comau S.p.A. The authors would like to thank Giovanni Di Stefano, Simone Panicucci, Nicola Longo, Luca Di Ruscio, Luca Robbiano, and Simone Peirone for their valuable assistance. This study was carried out within the FAIR - Future Artificial Intelligence Research and received funding from the European Union Next-GenerationEU (PIANO NAZIONALE DI RIPRESA E RESILIENZA (PNRR) – MISSIONE 4 COMPONENTE 2, INVESTIMENTO 1.3 – D.D. 1555 11/10/2022, PE00000013). This manuscript reflects only the authors’ views and opinions, neither the European Union nor the European Commission can be considered responsible for them. We thank the anonymous reviewers and editor for their constructive comments and suggestions.

Citation

If you find this work or code useful for your research, please consider citing our paper: 

@ARTICLE{11631852,
  author={Ortuno-Chanelo, Stephany and Rabino, Paolo and Civitelli, Enrico and Tommasi, Tatiana and Camoriano, Raffaello},
  journal={IEEE Robotics and Automation Letters}, 
  title={MultiGraspNet: A Multitask 3D Vision Model for Multi-Gripper Robotic Grasping}, 
  year={2026},
  volume={11},
  number={9},
  pages={10919-10926},
  keywords={Grasping;Grippers;Modeling;Multitasking;Learning (artificial intelligence);End effectors;Training;Switches;Weighted sum model;Clouds;Robotic grasping;multi-gripper grasping;3D perception;multitask learning},
  doi={10.1109/LRA.2026.3719235}}