Goal-Auxiliary Actor-Critic for 6D Robotic Grasping with Point Clouds

Lirui Wang, Yu Xiang, Wei Yang, Arsalan Mousavian, Dieter Fox
Proceedings of the 5th Conference on Robot Learning, PMLR 164:70-80, 2022.

Abstract

6D robotic grasping beyond top-down bin-picking scenarios is a challenging task. Previous solutions based on 6D grasp synthesis with robot motion planning usually operate in an open-loop setting, which are sensitive to grasp synthesis errors. In this work, we propose a new method for learning closed-loop control policies for 6D grasping. Our policy takes a segmented point cloud of an object from an egocentric camera as input, and outputs continuous 6D control actions of the robot gripper for grasping the object. We combine imitation learning and reinforcement learning and introduce a goal-auxiliary actor-critic algorithm for policy learning. We demonstrate that our learned policy can be integrated into a tabletop 6D grasping system and a human-robot handover system to improve the grasping performance of unseen objects. Videos and code are available at https://sites.google.com/view/gaddpg.

Cite this Paper

BibTeX
@InProceedings{pmlr-v164-wang22a, title = {Goal-Auxiliary Actor-Critic for 6D Robotic Grasping with Point Clouds}, author = {Wang, Lirui and Xiang, Yu and Yang, Wei and Mousavian, Arsalan and Fox, Dieter}, booktitle = {Proceedings of the 5th Conference on Robot Learning}, pages = {70--80}, year = {2022}, editor = {Faust, Aleksandra and Hsu, David and Neumann, Gerhard}, volume = {164}, series = {Proceedings of Machine Learning Research}, month = {08--11 Nov}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v164/wang22a/wang22a.pdf}, url = {https://proceedings.mlr.press/v164/wang22a.html}, abstract = {6D robotic grasping beyond top-down bin-picking scenarios is a challenging task. Previous solutions based on 6D grasp synthesis with robot motion planning usually operate in an open-loop setting, which are sensitive to grasp synthesis errors. In this work, we propose a new method for learning closed-loop control policies for 6D grasping. Our policy takes a segmented point cloud of an object from an egocentric camera as input, and outputs continuous 6D control actions of the robot gripper for grasping the object. We combine imitation learning and reinforcement learning and introduce a goal-auxiliary actor-critic algorithm for policy learning. We demonstrate that our learned policy can be integrated into a tabletop 6D grasping system and a human-robot handover system to improve the grasping performance of unseen objects. Videos and code are available at https://sites.google.com/view/gaddpg.} }
Endnote
%0 Conference Paper %T Goal-Auxiliary Actor-Critic for 6D Robotic Grasping with Point Clouds %A Lirui Wang %A Yu Xiang %A Wei Yang %A Arsalan Mousavian %A Dieter Fox %B Proceedings of the 5th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2022 %E Aleksandra Faust %E David Hsu %E Gerhard Neumann %F pmlr-v164-wang22a %I PMLR %P 70--80 %U https://proceedings.mlr.press/v164/wang22a.html %V 164 %X 6D robotic grasping beyond top-down bin-picking scenarios is a challenging task. Previous solutions based on 6D grasp synthesis with robot motion planning usually operate in an open-loop setting, which are sensitive to grasp synthesis errors. In this work, we propose a new method for learning closed-loop control policies for 6D grasping. Our policy takes a segmented point cloud of an object from an egocentric camera as input, and outputs continuous 6D control actions of the robot gripper for grasping the object. We combine imitation learning and reinforcement learning and introduce a goal-auxiliary actor-critic algorithm for policy learning. We demonstrate that our learned policy can be integrated into a tabletop 6D grasping system and a human-robot handover system to improve the grasping performance of unseen objects. Videos and code are available at https://sites.google.com/view/gaddpg.
APA
Wang, L., Xiang, Y., Yang, W., Mousavian, A. & Fox, D.. (2022). Goal-Auxiliary Actor-Critic for 6D Robotic Grasping with Point Clouds. Proceedings of the 5th Conference on Robot Learning, in Proceedings of Machine Learning Research 164:70-80 Available from https://proceedings.mlr.press/v164/wang22a.html.

Related Material