ALOHA Unleashed and DemoStart
Engineers from Google’s DeepMind project have announced the development of two new artificial intelligence-based robotic systems. A product called ALOHA Unleashed was developed to advance the science of dual-arm manipulation. Another, called DemoStart, aims to improve the functionality of robotic hands with multiple fingers, joints or sensors.
Details of ALOHA Unleashed have been posted on the DeepMind website and GitHub. Details of DemoStart have been posted on the arXiv preprint server.
As the research team points out, most robotic hands developed to pick up and move objects usually act alone-they have no second hand to help them. In this new effort, the research team used artificial intelligence technology to teach robots to use both hands together to complete a “difficult” task, such as tying shoelaces. The result is the “Aloha Release”
The team also pointed out that the new system is built on top of ALOHA 2 and the ALOHA platform, which was developed by Stanford University for remote operation of applications. The new system increases flexibility and also allows two robotic hands to be “aware” of each other when working together to solve common problems.
Demonstrations teach robot hands to perform tasks such as hanging shirts or repairing robot parts. Afterwards, diffusion methods were applied to provide robot hands with a certain degree of prediction, helping them predict what the other party would do.
DemoStart’s research team points out that the complex flexibility of the robot means it uses more fingers, joints and sensors than most current robotic hands. Achieving this goal requires some degree of coordination between them.
Like the ALOHA Unleashed project, coordination requires introducing artificial intelligence into the learning process. With DemoStart, they use reinforcement learning to help the robot understand its ability to control multiple arms, hands, and finger joints in addition to fingertips.
The method involves giving the robot hand simple tasks and then slowly increasing the difficulty. They found they could teach a two-fingered robot with multiple joints and sensors to reposition the cube, tighten nuts and organize the workspace.
There are links in the video below this video. If you are interested, you can open it and have a look.
Thank you for watching this video. If you like it, please subscribe and like it. thank
Original text:https://techxplore.com/news/2024-09-google-deepmind-unveils-ai-based.html
More information: Maria Bauza et al., DemoStart: Demonstration-led automated courses applied to simulation to reality of multi-fingered robots, arXiv (2024). DOI：10.48550/arxiv.2409.06613
ALOHA Unleashed: A simple secret to robot agility, aloha-unleashed.github.io/asse…/aloha_unleashed.pdf
DeepMind Blog: deepmind.google/discover/blog/…-in-robot-dexterity/
Start the demonstration: sites.google.com/view/demostart
Journal information: arXiv
Oil tubing: