• Skip to primary navigation
  • Skip to main content
  • Skip to primary sidebar
  • Skip to footer

My TechDecisions

  • Best of Tech Decisions
  • Topics
    • Video
    • Audio
    • Mobility
    • Unified Communications
    • IT Infrastructure
    • Network Security
    • Physical Security
    • Facility
    • Compliance
  • RFP Resources
  • Resources
  • Podcasts
  • Project of the Week
  • About Us
    SEARCH
News, Video

DeepMind Develops New Method to Transfer Simulated Skills to Robotic Arm

Artificial intelligence researchers at DeepMInd have developed a new method using reinforcement learning to teach a robot to stack blocks.

October 30, 2019 Zachary Comeau Leave a Comment

Reinforcement Learning
Featured Video Play Icon

Artificial intelligence researchers are improving the technology every day, including at Google parent company, Alphabet, where researchers have taught a robotic arm to stack blocks using reinforcement learning.

According to Venture Beat, the task has previously thought to be nearly impossible to do using only data.

DeepMind, Alphabet’s artificial intelligence research team, developed a new learning model that is like the work of OpenAI, which recently published a paper on methods of transferring skills from agents in simulation to a real-world robot to solve a Rubik’s cube.

DeepMind proposes a two-step procedure for adaption:

  1. In the first step, the team used a simulated environment to learn a policy that solves the cube-stacking task via synthetic images and proprioception. A state-based agent with access to the simulator and a vision-based agent that uses raw pixel observations train simultaneously. The state-based agent provides data for reinforcement learning of the vision-based agent.
  2. For the second step, unlabeled real image sequences help adapt the state representation to the real domain, which provides a common objective that applies to both simulation and reality. The DeepMind team says it mitigates the negative effects of the gap between simulation and the robot by leveraging unlabeled data collected by the simulation-trained agent.

DeepMind said their learning method yielded a clear improvement over domain randomization and other self-supervised adaption techniques, resulting in a cube-stacking success rate of 62% compared to the baseline 12% success rate. The technology could be applied to a wide range of robotic manipulation tasks, DeepMind says.

Do you know anyone that would fail at stacking cubes 38% of the time?

Maybe only if they’ve had a few. While it is impressive that AI researchers can get a robot to do mundane tasks without controlling it is an enormous achievement with reinforcement learning, the field of AI robotics has a long way to go.

In the OpenAI study, its robot, Dactyl, was able to solve a Rubik’s Cube with one robotic hand with what appears to be human-like dexterity.

Read Next: Artificial Intelligence Program Helps Fold Proteins

However, Dactyl need some serious work on its motor skills, as it dropped the cube eight out of 10 times in testing.

“I wouldn’t say it’s total hype—it’s not,” Ken Goldberg, a roboticist at UC Berkeley, told Wired. “But people are going to look at that video and think, ‘My God, next it’s going to be shuffling cards and other things,’ which it isn’t.”

If you enjoyed this article and want to receive more valuable industry content like this, click here to sign up for our digital newsletters!

Tagged With: Artificial Intelligence, DeepMind, Google, Robotics

Related Content:

  • Cloud, SASE, Aryaka How the Cloud is Redefining Media Production and…
  • Singlewire Software mass notification interview Singlewire Software on Mass Notification Solutions
  • URI catchbox 1 Catchbox Plus: The Mic Solution That Finally Gave…
  • Engaging virtual meeting with diverse participants discussing creative ideas in a bright office space during daylight hours Diversified Survey: Workplace AV Tech is Falling Short,…

Free downloadable guide you may like:

  • Practical Design Guide for Office SpacesPractical Design Guide for Office Spaces

    Recent Gartner research shows that workers prefer to return to the office for in-person meetings for relevant milestones, as well as for face-to-face time with co-workers. When designing the office spaces — and meeting spaces in particular — enabling that connection between co-workers is crucial. But introducing the right collaboration technology in meeting spaces can […]

Reader Interactions

Leave a Reply Cancel reply

Your email address will not be published. Required fields are marked *

Latest Downloads

Practical Design Guide for Office Spaces
Practical Design Guide for Office Spaces

Recent Gartner research shows that workers prefer to return to the office for in-person meetings for relevant milestones, as well as for face-to-fa...

New Camera Can Transform Your Live Production Workflow
New Camera System Can Transform Your Live Production Workflow

Sony's HXC-FZ90 studio camera system combines flexibility and exceptional image quality with entry-level pricing.

Creating Great User Experience and Ultimate Flexibility with Clickshare

Working and collaborating in any office environment today should be meaningful, as workers today go to office for very specific reasons. When desig...

View All Downloads

Would you like your latest project featured on TechDecisions as Project of the Week?

Apply Today!

More from Our Sister Publications

Get the latest news about AV integrators and Security installers from our sister publications:

Commercial IntegratorSecurity Sales

AV-iQ

Footer

TechDecisions

  • Home
  • Welcome to TechDecisions
  • Contact Us
  • Comment Guidelines
  • RSS Feeds
  • Twitter
  • Facebook
  • Linkedin

Free Technology Guides

FREE Downloadable resources from TechDecisions provide timely insight into the issues that IT, A/V, and Security end-users, managers, and decision makers are facing in commercial, corporate, education, institutional, and other vertical markets

View all Guides
TD Project of the Week

Get your latest project featured on TechDecisions Project of the Week. Submit your work once and it will be eligible for all upcoming weeks.

Enter Today!
Emerald Logo
ABOUTCAREERSAUTHORIZED SERVICE PROVIDERSYour Privacy ChoicesTERMS OF USEPRIVACY POLICY

© 2025 Emerald X, LLC. All rights reserved.