Robosobak taught to protect the football gate from the ball

Researchers from the University of California in Berkeley used reinforcement training to train the four -legged Mini Cheetah robot play football at the goalkeeper’s position.

“The task requires the implementation of highly dynamic movements with accurate and quick manipulations with a lack of object. The robot must react and intercept the potentially flying ball using various maneuvers in very short intervals, usually less than one second, ”the study says.

The device is able to perform three main movements: shift to the side, jump and dive. By the latter, scientists mean a sharp movement of the robot to the side with the subsequent fall to intercept the ball flying into the lower corner of the gate.

Mini Cheetah analyzes the position of the object using the depth chamber, after which it determines the right movement. AI-algorithm selects the trajectory of the movement of the robot parts that allows you to intercept the ball, and the low-level controller calculates the corresponding torque for the engines to realize the required movements.

First, the robot was trained to protect the gate in the ISAAC Gym Simulator from Nvidia. The platform allows you to implement training with reinforcement in a virtual space that imitates physical interactions.

Then the researchers transferred the acquired skills to the real Mini Cheetah and checked its effectiveness in playing football with a person and another robot. As a result, the device was able to successfully block the balls, spending 0.9 seconds on movements.

According to scientists, technology can be used to train the robot to play https://gagarin.news/news/will-the-new-digital-euro-protect-user-privacy/ football as an attacker.

“We focused exclusively on the task of creating a goalkeeper, but the proposed framework can be expanded to other scenarios, including the execution of the ball,” the study says.

Recall that in October, a team of engineers developed an algorithm system that allows robops to move in the wild. The devices equipped with the program can walk and run around a complex area, enveloping static and moving obstacles.

In October 2021, scientists from Bristol University took up robots to interact with radioactive waste.

In August, Boston Dynamics trained the two -legged Atlas robots overcome the obstacle strip and make a synchronous somersault back.

Subscribe to FORKLOG news at Telegram: Forklog AI – all news from the world of AI!