On October 28, according to foreign media reports, scientists trained a four-legged robot to become a goalkeeper. The quadruped robot goalkeeper was trained by scientists at UC Berkeley's Hybrid Robotics Laboratory.
Video footage shows it crouching, jumping, sideways and diving to block shots, and then returning to its starting position after blocking. starting position.
Experts say it can save 87.5% of shots on goal, compared to the average human goalkeeper's shot rate of about 69%.
Reinforcement learning is a subset of machine learning that allows artificial intelligence-driven systems to learn through trial and error using feedback from their behavior. This feedback is either negative or positive, in the form of punishment or reward, of course, with the goal of maximizing the reward function.
The four-legged robot was trained using reinforcement learning, a subset of machine learning that allows an artificial intelligence to learn through trial and error using feedback from its behavior.
"We propose a reinforcement learning framework that enables a quadruped robot to perform soccer goalkeeping tasks in the real world," the researchers said.
Football goalkeeping using a quadruped is a challenging problem that combines highly dynamic movements with precise and rapid manipulation of a non-prehensile object (ball).
"The robot needs to react to and intercept the ball that may fly out through dynamic movement actions in a very short period of time."
They then deployed the strategies it learned in Mini Cheetah, a quadruped robot developed at MIT and tested its performance in the real world. The results show that the reinforcement learning framework can greatly improve Mini Cheetah's ability as a soccer goalkeeper. In 40 random shots tested in the real world, the robot's save success rate reached 87.5%.
The above is the detailed content of The quadruped robot successfully defended the target with an interception success rate of 87.5%.. For more information, please follow other related articles on the PHP Chinese website!