New research by Professor Ken Goldberg, Jeff Mahler, and the Laboratory for Automation Science and Engineering (AUTOLAB) shows real progress on the problem of robots grasping everyday objects. DexNet 2.0 used deep learning with a cloud database of thousands of 3D objects to collect 6.7 million data points in order to train a robot to pick up and move objects in the real-world with a 99% success rate — significantly higher than previous methods.
With such a high success rate, it is likely that this work will soon be applied in industry, possibly revolutionizing manufacturing and the supply chain.
You can learn more about the research by reading the feature story at MIT Technology Review.
The research was the work of Jeffrey Mahler, Jacky Liang, Sherdil Niyaz, Michael Laskey, Richard Doan, Xinyu Liu, Juan Aparicio Ojea, and Ken Goldberg with support from the AUTOLAB team. The complete paper will be published in July.
UC Berkeley’s AUTOLAB, directed by Professor Ken Goldberg, is a world-renowed center for research in robotics and automation sciences, with 30+ postdocs, PhD and undergraduate students pursuing projects in Cloud Robotics, Deep Reinforcement Learning, Learning from Demonstrations, Computer Assisted Surgery, Automated Manufacturing, and New Media Artforms. Sponsors Include: NSF, USDA, DARPA, Google, Siemens, Intuitive Surgical, Autodesk, Samsung, Cisco, IBM, and CloudMinds. AUTOLAB Research Papers: http://goldberg.berkeley.edu/p