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

The article presents Any2Any, a novel paradigm for efficient cross-embodiment transfer of whole-body tracking (WBT) models in humanoid robots. It introduces techniques such as kinematic alignment and parameter-efficient fine-tuning (PEFT) to adapt pretrained models to new embodiments using only 1% of the data and compute typically required for full training. This approach significantly reduces deployment costs and training time while maintaining or improving tracking performance, offering a scalable solution for implementing WBT across different humanoid platforms.