AI Trains Miniature Robots to Play Football

British scientists have pushed the envelope in robotics by enabling tiny humanoid robots—just knee-high to a person—to play football. These mini-athletes are powered by deep reinforcement learning, a cutting-edge AI technology that assimilates vast amounts of football simulation data to equip the robots with versatile in-game tactics.

These robots can autonomously choose tactics, dynamically adapt to game situations, and even fiercely compete for the ball. While currently limited to one-on-one matches, their movement speed, turning, and ball-striking have surpassed initial expectations. They’ve also achieved smoother, more precise movements and can automatically determine their tactics almost as quickly as live players do, blocking shots and standing up after falls.