Robots created by the MIT Biomimetic Robotics Lab are being taught how to play soccer at University of California Berkeley.
In 2019, a group of soccer players were seen kicking around a ball on MIT’s Killian Court lawn. They ran around, jumped in piles of leaves, and even did backflips. But they weren’t students. They were Mini Cheetah soccer-playing robots that are powered by 12 motors. Three years later, researchers from the Hybrid Robotics Lab at the University of California, Berkeley are teaching similar four-legged robots how to be a goalie.
The soccer-playing robot was developed at MIT Biomimetic Robotics Lab and was taught how to goal keep through simulation. During several tests, the robot invention managed to pull off a pretty strong blocking average and saved 87.5% of the shots taken. This is just slightly above that of the best professional goalkeepers in the English Premier League. While it seems like a fun exercise, the applications are rather significant.
Teaching mini cheetah robots to play soccer helps to solve broader challenges within robotics. Authors of a new paper detailing the use of reinforcement learning to teach MIT’s soccer-playing robot to keep goal explained the complexities of the movement. “Soccer goalkeeping using quadrupeds is a challenging problem that combines highly dynamic locomotion with precise and fast non-prehensile object (ball) manipulation,” they said.
The soccer-playing robot needs to react to, and intercept a potentially flying ball using dynamic locomotion maneuvers in a very short amount of time – usually less than one second. In the paper, the scientists propose to address the problem using a hierarchical model-free RL framework. Effectively, the robot needs to lock into a projectile and maneuver itself to block the ball in under a second, TechCrunch reports.
The soccer-playing robot’s parameters are defined in an emulator. And the Mini Cheetah relies on a trio of moves, which includes sidestepping, diving, and jumping, to block the ball on its way to the goal by determining its trajectory while in motion. To test the efficacy of the robotics program, the research team pitted the system against a human component and a fellow Mini Cheetah. Interestingly, the same basic framework used to defend the goal area can be applied to offense.
The paper’s authors note, “In this work, we focused solely on the goalkeeping task, but the proposed framework can be extended to other scenarios, such as multi-skill soccer ball kicking.” The team is hopeful that their soccer-playing robots could eventually help with anything that requires a human being to travel a distance and then do a specific physical action. This includes things like deliveries, elder care, and emergency response.
But this isn’t the first time scientists have tinkered with soccer-playing robots. Earlier this year, a team of researchers at Google’s Deep Mind London project taught animated players how to play a realistic version of soccer on a computer screen. In their paper, published in the Science Robotics journal, the group describes teaching the animated players to play as solo players and also in teams.
The idea behind the research is to get simulated soccer-playing robots to learn the game by watching others do it. This is the same way humans learn to play soccer. The simulated players first had to learn how to walk, then to run and kick a ball around, TechXplore reports. At each new level, the AI systems were shown a video of real-world soccer players, which allowed them to learn not just the basics of the game but to mimic the way professional athletes move.