CloudGripper: The Open Source Cloud Robotics Testbed

Introduction

CloudGripper is an open source cloud robotics testbed for remote robotic manipulation research, benchmarking and data collection at scale with a current deployment of 32 robot arm cells at KTH Royal Institute of Technology, Stockholm, Sweden. The aim of the project is to create a global community of contributors and remote users of this testbed that will leverage CloudGripper for research, benchmarking and data collection. Come visit us at our booth at IEEE ICRA in Yokohama, Japan to find out more.

Newsletter, Signup and Contact

You can sign up to our newsletter here. Alternatively, contact us at info (at) cloudgripper (dot) org.

Architecture description

A preprint about CloudGripper is available here: "CloudGripper: An Open Source Cloud Robotics Testbed for Robotic Manipulation Research, Benchmarking and Data Collection at Scale" (accepted at IEEE ICRA 2024) by Muhammad Zahid and Florian T. Pokorny.

Open Software and Hardware

You can find sofware and CAD files under open source licenses on our GitHub organization: CloudGripper on GitHub. Assembly instructions are available here. A MuJoCo simulation of CloudGripper will also be made available shortly. The CloudGripper-Autograsper software can be found here together with an associated dataset here.

Large Open Datasets

We are currently finalizing the best options for hosting the large multi terabyte datasets being generated with the CloudGripper robots. At the moment two datasets of a total of approximately 2 terabytes of randomized pushing interactions are available. A data sample from the CloudGripper-Rope-100 dataset of more than 100 hours of rope pushing interactions is available here. A further datasample of our recently submitted work "Video Transformers under Occlusion: How Physics and Background Attributes Impact Large Models for Robotic Manipulation" Shutong Jin, Ruiyu Wang, Muhammad Zahid, Florian T. Pokornywhich resulted in a video pushing dataset with more than 1000 hours of data can be found here.

Please email info (at) cloudgripper (dot) org for access to the full dataset and more information.

Current Research Projects using CloudGripper

The initial release of CloudGripper was developed by Muhammad Zahid and Associate Prof. Florian Pokorny at the Division of the Robotics, Perception and Learning at KTH Royal Institute of Technology. A cloud robotics test-bed installation based on this design is currently being scaled with funding for more than 100 robots as part of the WASP funded Swedish excellent research environment (NEST) grant "Intelligent Cloud Robotics for Real-Time Manipulation at Scale" coordinated by Florian T. Pokorny and with co-investigators Monowar Bhuyan, Erik Elmroth and Martina Maggio. This work is partially supported by the Wallenberg AI, Autonomous Systems and Software program (WASP) funded by the Knut and Alice Wallenberg Foundation.

CloudGripper community members

We are building a community around cloud robotics. Are you interested in joining us and use the system? Let us know.