Some robotics challenges have immediately obvious applications. Others focus more on helping systems solve larger problems. Teaching little robots to play football against each other falls squarely into the latter category.
The authors of a new paper detailing the use of reinforcement learning to teach MIT’s Mini Cheetah robot to play goalie note:
The football goalkeeper using quadrupeds is a difficult problem that combines highly dynamic locomotion with precise and rapid manipulation of a non-prehensile object (ball). The robot must react and intercept a potentially flying ball using dynamic locomotion maneuvers in a very short time, typically less than a second. In this article, we propose to solve this problem by using an RL framework without a hierarchical model.
In effect, the robot must lock into a projectile and maneuver to block the bullet in less than a second. Robot settings are set in an emulator, and the Mini Cheetah relies on a trio of moves – dodging, dipping and jumping – to block the ball on its way to the goal by determining its trajectory in motion.
To test the effectiveness of the program, the team pitted the system against both a human component and another Mini Cheetah. Notably, the same basic framework used in defending the goal can be applied in attack. The authors of the article note: “In this work, we have focused only on the goalkeeper task, but the proposed framework can be extended to other scenarios, such as the multi soccer ball kick. -skills.”