DynaRetarget: Dynamically-Feasible Retargeting using Sampling-Based Trajectory Optimization

2026-02-06

AI Summary

This work addresses the challenge of dynamically infeasible motion retargeting for humanoid robots performing long-horizon tasks involving diverse object interactions. To this end, the authors propose DynaRetarget, a framework that integrates sampling-based trajectory optimization (SBTO) with an incremental optimization horizon to progressively refine kinematic demonstrations into dynamically feasible full-body motions within physical simulation. A key innovation lies in its ability to generalize across objects of varying mass, size, and geometry without requiring adjustments to the tracking objective, while enabling global dynamic optimization over entire motion trajectories. Experimental results demonstrate that DynaRetarget successfully retargets hundreds of loco-manipulation demonstrations with significantly higher success rates than existing methods, offering an effective pathway toward constructing large-scale synthetic datasets for humanoid robotics.

Technology Category

• Intelligent Robots: Manipulation
• Search and Optimization: Sampling/Simulation-based Search
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

In this paper, we introduce DynaRetarget, a complete pipeline for retargeting human motions to humanoid control policies. The core component of DynaRetarget is a novel Sampling-Based Trajectory Optimization (SBTO) framework that refines imperfect kinematic trajectories into dynamically feasible motions. SBTO incrementally advances the optimization horizon, enabling optimization over the entire trajectory for long-horizon tasks. We validate DynaRetarget by successfully retargeting hundreds of humanoid-object demonstrations and achieving higher success rates than the state of the art. The framework also generalizes across varying object properties, such as mass, size, and geometry, using the same tracking objective. This ability to robustly retarget diverse demonstrations opens the door to generating large-scale synthetic datasets of humanoid loco-manipulation trajectories, addressing a major bottleneck in real-world data collection.

Problem

• motion retargeting
• dynamic feasibility
• humanoid locomotion
• trajectory optimization
• data collection bottleneck

Innovation

• Sampling-Based Trajectory Optimization
• Motion Retargeting
• Dynamic Feasibility
• Humanoid Loco-Manipulation
• Long-Horizon Optimization