caption

Breakthrough AI Robot Tech JUST Learned THIS Better Than OpenAI + Google With Reinforcement Learning

Artificial intelligence powered robots just learned how to grip even ungraspable
objects using an amazing technique. Carnegie Mellon University's Robotics Institute has demonstrated that even basic
robotic rippers can carry out complicated tasks by taking advantage
of their surroundings. These grippers are usually only used for simple tasks, such as
picking up and placing objects. However, by using their surroundings,
such as pushing an object checked against a wall or table, they are able to perform
actions that were previously thought to be possible only with more advanced and
costly multifingered artificial hands. Previous research on this method, called extrinsic dexterity,
often made assumptions about how grippers would grasp items and required specific
gripper designs or robot movements. Instead, these robotics researchers made this
breakthrough by using artificial intelligence and the technique
of extrinsic dexterity to grasp objects of various sizes, shapes,
weights and surfaces. Here's how they did it: first, they trained a neural network using
reinforcement learning, where the AI system was rewarded
for successful attempts at grasping an object to learn the most effective
patterns of behavior through trial and error. The robotics researchers
initially trained the system in a physics simulator and then tested it on a simple
robot with a pincer-like grip.

Next, the researchers placed objects
in a bin in positions that made it difficult for the robot
to grab them directly. For instance, the robot might be given an object that was too wide
to be grasped by its gripper. Through reinforcement learning, it had to figure out how to push
the object against the wall of the bin so that the robot could then
grab it from the side. The robotics researchers initially thought
the robot might try to use a scooping motion to pick up the object,
similar to how humans do it. However, the artificial intelligence algorithm came up with
a different solution. After pushing the object against the wall,
the robot used its top finger to lift the object and then let it fall
into the bottom finger to grasp it.

The robotic system was tested on various
objects, including cardboard boxes, plastic bottles, a toy purse,
and a container of Cool Whip. These objects varied in weight,
shape and slipperiness. They found that basic grippers were able to successfully grasp these
items 78% of the time. The researchers hope to expand their findings to include a wider range
of objects and scenarios in the future. They are also interested in exploring more
complex tasks that can be performed with a basic gripper using their
extrinsic dexterity technique. But where could this robotic gripping and manipulation technology
be applied in the real world? Enter the Aeo, a humanoid AI robot that could make direct
use of this kind of artificial intelligence driven gripping
technology from Aeolus Robotics. The Aeo robot was just unveiled at CES 2023
as an autonomous humanoid robot with a dual arm design for delivery,
elder care, security services, and more. The Aeo robot uses advanced deep learning, artificial intelligence and sensing
to adapt to its surroundings and the people nearby, ensuring
safety and efficiency for everyone, and due to its onboard AI, the robot improves and learns over time
as it interacts with its environment.

The Aeo has two robotic arms,
one of which is used to carry out important tasks such as delivery or
disinfection, and the other operates elevators and opens doors
without any modifications. Aeo also uses its artificial intelligence
to dynamically navigate and avoid obstacles while reacting
to people in real time. The robot also has additional capabilities
such as kiosk operation and disinfection using ultraviolet light,
as well as the ability to open doors and ride elevators to extend
its range of movement. It can be customized with plug and play
attachments from third party developers such as Asratec in Japan
and Malibu AI in Taiwan. Aeolus has emphasized that Aeo's dual arms
have seven degrees of freedom, allowing the robot to perform
different tasks with each hand. Aoelus Robotics says their company's aim is to develop robots that will improve
the quality of life and further the positive impact that robots
already have on society. In 2018, they released their first single
arm robot, which set a high standard for what a robotic
assistant could accomplish. However, the company wanted to do more. Their experience deploying robots
in various settings such as senior care facilities, hospitals,
public transportation offices and hotels is what motivated them to create
the dual arm humanoid robot.

So far, the Aeo humanoid AI robot has already been deployed in Japan,
Hong Kong and Taiwan where it provides services for major elder
care providers and real estate companies. Aeolus Robotics also helps operate a hotel with a timeshare membership
that includes a senior care department that combines hospitality,
medical care and nursing care. The robot is already a member of the team and plays a crucial role by completing
simple tasks that allow care facility staff to have more time to interact
with patients and staff. As a result, Aeolus Robotics expects both
resident and patient care interaction as well as hospitality services
to improve breakthrough.

Leave a Reply

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