An American robotics company has used artificial intelligence (AI) to give their humanoid robots to give more natural looking strid, and they say it is just the beginning.
In Publicity videoRobot, called Figure 02 and marches with “human-like” moves, manufactured by the company’s figure. This is a potential that claims that its robot will help the physical world to navigate more easily.
“These initial results are exciting, but we believe that they only indicate the full potential of our technology,” the representative of the company Written in a blog post With its announcement. “We are committed to increasing our learned policy to handle every human-like landscape that can cope with the robot in the real world.”
Are robots run, Cartaviling, Backflip, break dancingAnd we have been beating us in chess for years now. But such tasks that appear simple for humans, such as Walking in a straight line, Catch items, Shool, And navigating social conditions, has proved to be difficult to master for robots.
Connected: Chinese Humanoid Robot Sneakers ‘Fast’ for the trusted pair of sneakers
The problem is known as Morveak’s contradictionEmerges because computers excel on problems that require complex calculations and large datasets, but lack millions of years of development and lack our real world experience. It changes robots, well, robot best. The worst, it gives them the appearance that they can dirty themselves.
To deal with the unnatural move of the robot, the engineers of the figure used a learning technique called reinforcement – a physics simulator placed thousands of virtual robots inside a physics simulator that rebuilds different areas, which improves their walking through testing and error.
By rewarding the virtual robot army for natural movements, he refined his gates to look more human-like. Along with completing this task, he uploaded the refined “learned natural foot” model for a real -world images 02 robots. The result is an Android that can some extent move naturally, with heel attacks, toe-closes and synchronized arm swing.
Figure’s reinforcement learning technique is important for California company’s plans to roll your robot on the floor of the factory. It has already tested its humanoid robot in a BMW factory in 2024 and is planning to present more this year. Meanwhile, a competitive Apptronik, a Texas-based competitive, is commercializing its humanoid robots, Apollo for use in Mercedes-Benz factories by the end of 2025. The figure of agility robotics will also be introduced in the warehouse this year.
