Soccer-playing humanoid robot learns four skills from human motion data

A single control policy enables a robot to move, dribble, trap, and kick with human-like fluidity and zero-shot transfer to a real robot.

Top University

Haoyu Shao · Yunshen Wang · Jingze Zhang · Shaohang Zhu · Yong-Lu Li · Tianzhu Zhang · +2 more

University of Science and Technology of China · State Key Laboratory of General Artificial Intelligence, BIGAI · Shanghai Jiao Tong University · Peking University

Research Digest·5 November 2026·2 min read

Illustration: Soccer-playing humanoid robot learns four skills from human motion data

In one line

SoccerBot is a humanoid soccer system that uses human motion data to learn a single policy for moving, dribbling, trapping, and kicking with human-like motion.

Before you read this

An idea this paper builds on

reinforcement learningA machine learning method where an agent learns to make sequences of decisions by receiving rewards or penalties for its actions.261 papers →

Read it as

Playing soccer requires a robot to combine running, dribbling, trapping, and kicking with natural movement. Existing solutions either cover only a few skills or produce unnatural motions.

The researchers developed SoccerBot, which first learns a 'motion prior' from recordings of humans playing soccer. This prior captures natural movement patterns. Then a single control policy is trained to use that prior to perform all four skills.

The system transferred directly to a real Unitree G1 humanoid robot without additional training, and executed all skills with agile, natural motion. This matters because it shows robots can learn versatile, human-like behaviors from data, advancing robot soccer and general robot control.

How it works

stage by stage

RSS

The Daily Edition

The front page in your inbox each weekday morning: the lead, the picks, and why each one earned its place.

Concepts

latent motion prior reinforcement learning sim-to-real transfer knowledge distillation domain randomization curriculum learning