Locomotion-Grounded Humanoid Soccer: Task-Gated Reinforcement Learning of a Multi-Directional Kicking Library

2026-09-29

AI Summary

This study addresses the challenges of control constraints and poor scalability of kicking skills in humanoid robot soccer by proposing a task-gated reinforcement learning framework driven by universal instruction-based locomotion policies. The core innovation lies in inverting the conventional technical stack, treating fundamental locomotion as the base layer upon which multi-directional kicking skills are efficiently superimposed with O(N) complexity. By integrating motion retargeting with instruction-conditioned training, the proposed approach successfully implements seven distinct kicking actions spanning 259.5 degrees—including non-conventional lateral and backward directions—on the Unitree G1 platform. This framework significantly enhances shooting accuracy and terrain adaptability, offering a scalable solution for complex skill acquisition in legged robotics.

Abstract

Recent humanoid soccer systems make motion tracking the substrate and derive locomotion from it, typically by steering a motion-reference anchor toward the ball. This yields strong shooting results, but locomotion is trained only on the narrow, deterministic command distribution ball approach induces, never evaluated as a capability in its own right. We invert the stack: a general, command-conditioned locomotion policy is trained first as the substrate, and N motion-guided kicking skills are added on top as task-gated layers, so the reachable gait space is set by the locomotion curriculum rather than any reference clip. Because every skill starts from and returns to this same commandable state, locomotion also becomes a composition hub (O(N) transitions rather than O(N^2)), and post-strike stabilisation is handed back to the trained controller rather than scripted per clip. We instantiate this on a 29-DoF Unitree G1 with seven retargeted kicking skills spanning 259.5 degrees of nominal aim direction, including lateral, rearward and weak-foot strikes a single forward-facing reference cannot express, and report shooting accuracy alongside command-tracking, terrain and push-recovery results with the full skill library attached, an axis prior humanoid soccer systems do not report. The library is validated on hardware across forward, lateral, rearward and commanded approaches.

Problem

• humanoid soccer
• locomotion
• multi-directional kicking
• reinforcement learning
• skill composition

Innovation

• Task-Gated Reinforcement Learning
• Command-Conditioned Locomotion
• Humanoid Soccer
• Multi-Directional Kicking
• Sim-to-Real Transfer