The robot already knows how to dribble, pass people, and score goals

Thanks to the new AI training method, the robots were able to automatically improve their skills and participate in a simulated soccer match .

Watch robots trained by AI dribble, pass people, and score goals. (Video: Google DeepMind).

Humanoid robots have long been a product of scientists' ambition to create a device that can simulate all human activities, thereby serving us for many different purposes.

However, the development of humanoid robots still faces many limitations, because every movement needs to be programmed. This requires scientists to synthesize huge amounts of data.

However, today's robot technology has received great help from AI. This creates a major turning point in developing and putting robots into practice.

AI's self-learning ability means that robots will not need to move according to existing programming, but are fully capable of performing techniques without direct human control.

In new research, scientists at Google DeepMind trained the Robotis OP3 robot to play soccer using a technique called "deep reinforcement learning".

This is a machine learning training technique that combines many different AI training methods. There, the robot uses an algorithm network that operates like artificial neurons, and is arranged similar to the human brain.

Robots trained by AI have developed emergency movement behaviors that are extremely difficult to program, such as turning the angle of the foot, spinning around to deceive the opponent.

In simulated matches, the AI-trained robot moved 181% faster, turned 302% faster, kicked a ball 34% faster and took 63% less time to recover from falls compared to with robots that are not trained to use this technique.

From the actual video recorded, it can be seen that the robot can perform complex movements such as dribbling, blocking, passing people, scoring goals. relatively skillfully and at high speed.

The researchers added that the robot has also developed emergency movement behaviors that are extremely difficult to program, such as rotating the angle of the foot, spinning around to deceive the opponent.

The findings suggest that this AI training technique can be used to create simple, but relatively safe, movements in humanoid robots in general. From this starting point, the robot can be enhanced with more complex movements, and applied in more practical situations.

The same technology applied also helped the robot have the ability to make coffee on its own after a training period of only 10 hours.