Any2Any: Efficient Cross-Embodiment Transfer for Humanoid Whole-Body Tracking

2026-05-22

AI Summary

This work addresses the high cost and slow deployment associated with training humanoid whole-body tracking models from scratch. The authors propose the Any2Any transfer paradigm, which leverages kinematic alignment and lightweight dynamic adaptation, combined with parameter-efficient fine-tuning (PEFT) and reuse of pretraining strategies, to effectively transfer high-performance tracking capabilities to new robot morphologies. Requiring only 1% of the original data and computational resources, this approach achieves convergence significantly faster across multiple robotic platforms while delivering performance comparable to or even surpassing that of models trained from scratch. By drastically reducing the resource overhead, the method substantially lowers the barrier to deploying advanced tracking systems on diverse humanoid robots.

Abstract

Whole-body tracking (WBT) models have become a key foundation for humanoid robots, enabling them to imitate diverse motions with high fidelity. Training such models from scratch requires large-scale data and computation, making rapid deployment on new humanoid platforms costly. This raises a natural question: Can pretrained WBT models transfer across embodiments with minimal adaptation? To answer this question, we propose Any2Any, a paradigm that efficiently transfers an existing WBT specialist to a new humanoid embodiment with only a small amount of data and compute. Any2Any first performs kinematic alignment between source and target humanoids, aligning their input and output spaces so that the pretrained source policy can be meaningfully reused on the target embodiment.Any2Any then performs dynamics adaptation by applying lightweight parameter-efficient fine-tuning (PEFT) components to selected dynamics-sensitive modules, preserving useful behavioral priors while enabling targeted adaptation to the target robot. Extensive experiments on multiple humanoid platforms and pretrained backbones show that Any2Any substantially accelerates convergence and reduces training cost compared with training from scratch, while achieving competitive or superior tracking performance. Notably, using only 1% of the compute and data required for full training, Any2Any successfully transfers Sonic models pre-trained on Unitree G1 to LimX Oli and LimX Luna. These results suggest that pretrained WBT specialists can be efficiently reused across embodiments, providing a scalable path toward deploying humanoid whole-body control on new robots.

Problem

• whole-body tracking
• cross-embodiment transfer
• humanoid robots
• model adaptation
• pretrained models

Innovation

• cross-embodiment transfer
• whole-body tracking
• kinematic alignment
• parameter-efficient fine-tuning
• humanoid robots