Tsinghua team lands in Science Robotics after 22 years of robot soccer work

A Tsinghua University team led by Professor Zhao Mingguo has published a paper in Science Robotics with ByteDance Seed and China Agricultural University on vision-driven reactive soccer skills for humanoid robots. The study uses an AgiBot humanoid platform for real-robot tests and covers ball finding, chasing, gait adjustment and multi-direction shooting. It also traces a 22-year line from the Tsinghua Vulcan robot soccer team, founded in 2004, to AgiBot’s current hardware platform and the team’s RoboCup work. The paper reports a unified perception-motion reinforcement learning framework, with visual sensing, state estimation, locomotion and physical contact trained together for a continuous soccer task. On the experiment side, the team says the robot can keep reacting when vision is briefly interrupted, and that it achieved stable results in both simulation and hardware tests. The article also notes AgiBot’s Booster T2 platform, unveiled in 2025, and its use in RoboCup and the Robot Competition opening ceremony in 2026.