Carnegie Mellon and NVIDIA University cooperated to develop a new training technology that enables human robots to perform complex sports movements with the unprecedented light movement-from Cristiano Ronaldo's distinguished celebration in the middle of the air to the Cuban Bryant leap shot.
Frame, The alignment of real simulation and physics (As quickly as possible), it embodies a critical gap between simulation and reality by allowing human robots to implement high -level sporting movements that have already been thoughtful in the past very complicated for machines.
In their paper, the researchers pointed out that "human robots bear the possibility of unparalleled diversity to perform human -like skills." "However, achieving full, aggravated and coordinated body movements is still a major challenge due to the lack of dynamics between simulation and the real world."
As soon as possible, this challenge is treated with a two -stage process.
First, it performs the movement tracking exercises before training-algorithm that controls the trace-simulation using human movement data. Then these policies are published in the real world to collect data that helps bridge the gap between simulator and actual physics.
The result is a human robot capable of Repeated moves signing from the legend of sportsSIncluding the famous "SIU" celebration of Cristiano Ronaldo (including 180 degrees in the air), and the "Sound Sound" party for Liberal James (features one minute) and budget Bryant's jump).
Besides these luxurious sporting moves, the robot showed other impressive exploits such as the front and side jumps of more than 1 meter.
At first glance, robots may still look peak, but this time, this is often due to the restrictions of the devices, because they have less human expression.
However, they have greater ingenuity than other robots thanks to the "Delta business model"-a correction mechanism compensated for the differences between simulation and realistic physics. "Delta's business model acts effectively as a term remaining correction of the dynamic gap."

Using this approach, the researchers reduced the tracking errors by up to 52.7 % compared to the previous methods, allowing robots to make complex movements that were previously impossible.
The researchers, who have shown that the effectiveness of the system "paves the way for multi -use human robots in the real world's applications."
The development of robots with this level of ingenuity is particularly difficult and was one of the most continuous challenges in robots.
"For decades, we imagined the human robots that achieve or even exceed the lightness of movement at the human level. However, most of the previous works focused primarily on the movement, and the treatment of legs as a means of transportation." The researchers wrote.
As soon as possible, on the other hand, the human body simulates pre -training and is able to adapt its knowledge to the real world scenarios after identifying it in simulation.
In this way, robot ends such as the human limbs, which are used for movement, balance, weight balance, expression and more.
Achieving this is much more difficult than it seems. When we perform athletic performance-even basic-we actually prevent countless accurate adjustments, and the balance of multiple forces with compensation for changes in momentum and position.
Robots have been proven to repeat this very difficult.
Do not believe us? Try to play Qwop- It is a game that you must control 4 arthritis to make a sporty. Now, as soon as you spend hours mastering that game, consider how difficult it is to manage the 21 basic expressions at the same time that it deals with as quickly as possible - then consider that the human body has more than 300 different joints.

It was the field of human robots In recent yearsWith companies and universities that pour more resources in research and development.
The Tesla's Optimus Project and Figure Humanoid Robot from AI and Boston Dynamics increased in human robots.
In the academic region, Bristol University and Stanford They also developed their own methods to teach models how to be more flexible and increase their ingenuity.
The team focuses on further development as soon as possible
They said, "Future trends can focus on developing policy structures that realize damage to alleviate the risk of devices," referring to how some models erupted while trying to make complex movements.
They also want to search "benefit from positivism without signs or integrate the sensor on the plane to reduce dependence on MCAP systems" and improve their adaptation techniques to achieve higher efficiency. How long before we have a robot World Cup?
Edit Sebastian Senkler and Josh Ketner
Smart in general Newsletter
Weekly journey narrated GEN, AI Tawylidi model.
Source link