Dribble Master: Learning Agile Humanoid Dribbling through Legged Locomotion
Zhuoheng Wang, Jinyin Zhou, Qi Wu
AI summary
Problem
Humanoid soccer dribbling demands precise ball manipulation while maintaining dynamic balance, yet traditional methods struggle with complex foot-ball interactions, terrain variability, and discontinuous visual perception.
Approach
The method decouples training into a locomotion stage and a dribbling fine-tuning stage, using a virtual camera model and heuristic rewards to teach the robot active ball tracking without predefined trajectories.
Key results
- Achieves 86.7% success rate in target dribbling and 93.3% in obstacle avoidance, surpassing teleoperation and ablation baselines
- Successfully deploys the learned policy on the physical Booster T1 humanoid robot for real-world dribbling
- Enables active sensing by learning to reorient its head to maintain ball visibility despite narrow camera fields of view
- Demonstrates agile directional changes and robust velocity tracking across varied terrains and friction conditions
Why it matters
This work demonstrates that curriculum learning and active sensing can overcome sim-to-real gaps, advancing practical legged loco-manipulation and athletic intelligence for humanoid robots.
Abstract
Humanoid soccer dribbling is a highly challenging task that demands dexterous ball manipulation while main- taining dynamic balance. Traditional rule-based methods often struggle to achieve accurate ball control due to their reliance on fixed walking patterns and limited adaptability to real-time ball dynamics. To address these challenges, we propose a two- stage curriculum learning framework that enables a humanoid robot to acquire dribbling skills without explicit dynamics or predefined trajectories. In the first stage, the robot learns basic locomotion skills; in the second stage, we fine-tune the policy for agile dribbling maneuvers. We further introduce a virtual camera model in simulation that simulates the field of view and 1This work was done before the author joined Cornell University. perception constraints of the real robot, enabling realistic ball perception during training. We also design heuristic rewards to encourage active sensing, promoting a broader visual range for continuous ball perception. The policy is trained in simulation and successfully transferred to a physical humanoid robot. Ex- periment results demonstrate that our method enables effective ball manipulation, achieving flexible and visually appealing dribbling behaviors across multiple environments. This work highlights the potential of reinforcement learning in developing agile humanoid soccer robots. Additional details and videos are available at https://zhuoheng0910.github.io/dribble-master/. 2026 IEEE International Conference on Robotics and Automation (ICRA 2026) June 1-5, 2026. Vienna, Austria 979-8-3315-8160-2/26/$31.00 ©2026 IEEE 5853