Iterative Closed-Loop Motion Synthesis for Scaling the Capabilities of Humanoid Control

2026-02-25

AI Summary

Existing physics-based humanoid control methods rely on motion datasets of fixed difficulty, which are constrained by the high cost and limited scalability of high-quality motion capture data. This work proposes a closed-loop framework for automatic motion data generation and iterative refinement, enabling co-evolution of control policies and training data through physics-aware simulation, semantically rich motion synthesis (e.g., martial arts and dance), difficulty-aware data iteration, and goal-directed evaluation. Using only approximately one-tenth the data volume of the AMASS dataset, the PHC single-motion tracker reduces the average failure rate by 45% across 2,201 test motion clips, substantially surpassing previous performance bottlenecks.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Computer Vision: Motion & Tracking
• Intelligent Robots: Motion and Path Planning

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Physics-based humanoid control relies on training with motion datasets that have diverse data distributions. However, the fixed difficulty distribution of datasets limits the performance ceiling of the trained control policies. Additionally, the method of acquiring high-quality data through professional motion capture systems is constrained by costs, making it difficult to achieve large-scale scalability. To address these issues, we propose a closed-loop automated motion data generation and iterative framework. It can generate high-quality motion data with rich action semantics, including martial arts, dance, combat, sports, gymnastics, and more. Furthermore, our framework enables difficulty iteration of policies and data through physical metrics and objective evaluations, allowing the trained tracker to break through its original difficulty limits. On the PHC single-primitive tracker, using only approximately 1/10 of the AMASS dataset size, the average failure rate on the test set (2201 clips) is reduced by 45\% compared to the baseline. Finally, we conduct comprehensive ablation and comparative experiments to highlight the rationality and advantages of our framework.

Problem

• humanoid control
• motion dataset
• data scalability
• difficulty distribution
• physics-based simulation

Innovation

• closed-loop motion synthesis
• iterative policy training
• physics-based humanoid control
• scalable motion generation
• difficulty-aware data iteration