Google DeepMind teaches robot to autonomously clean a kitchen, tie shoelaces, and other tasks
The advancements within robotics continue apace as Google DeepMind makes headway in dexterity, with complex tasks that require dexterous movement now possible.
Previously, AI robots have only been able to pick up and place objects using a single arm. The research lab’s new methods have meant their robot has “learned to tie a shoelace, hang a shirt, repair another robot, insert a gear and even clean a kitchen.”
Manipulating objects using just one robot arm is challenging.
Enter ALOHA Unleashed, which builds upon our ALOHA 2 system. With two arms, it can be teleoperated to collect high quality training data.
With this system, robots can perform new tasks with fewer demonstrations. pic.twitter.com/D6J0P3eqxJ
— Google DeepMind (@GoogleDeepMind) September 12, 2024
The team’s new advances include ALOHA Unleashed which helps robots learn to perform complex and novel two-armed manipulation tasks and DemoStart which introduces simulations to improve real-world performance on a multi-fingered robotic hand.
Since the development of robots first began, highly-dexterous tasks have long been considered the most difficult to get right. In the company’s latest blog, the team suggests it’s these tasks that would be most “useful in people’s lives.”
To achieve the new tasks, the team first collected “demonstration data by remotely operating the robot’s behavior, performing difficult tasks like tying shoelaces and hanging t-shirts.”
From there, a diffusion method was applied which predicted robot actions from random noise. This means the robot was able to learn from the data, so it could perform the same tasks on its own.
An AI controlling a multi-fingered robotic hand means it could carry out more useful, physical actions.
DemoStart takes us one step closer by using a reinforcement learning algorithm to master various behaviors from just a handful of simulated demonstrations. pic.twitter.com/2t28faw9jU
— Google DeepMind (@GoogleDeepMind) September 12, 2024
The robot had a 98% success rate on numerous different tasks in simulation, including reorienting cubes with a certain color showing, tightening a nut and bolt, and tidying up tools.
The future of robots according to Google DeepMind
While robotics is a unique area of AI research, there is work ongoing across multiple companies and startups to increase use cases.
In Google DeepMind’s blog about the update, the robotics team states they “still have a long way to go before robots can grasp and handle objects with the ease and precision of people.”
The researchers add they have made “significant progress” and call “each groundbreaking innovation” a “step in the right direction.”
“One day, AI robots will help people with all kinds of tasks at home, in the workplace and more. Dexterity research, including the efficient and general learning approaches we’ve described today, will help make that future possible.”
Featured Image: Via GoogleMind blog/credit to Shadow Robot
The post Google DeepMind teaches robot to autonomously clean a kitchen, tie shoelaces, and other tasks appeared first on ReadWrite.