# 持续人形机器人运动学习

> 原文链接：https://robotics.hellobike.com/papers/2026-10-03-2610-04231
> 频道：papers | 更新日期：2026-10-03 | 标签：arxiv, cs.RO
> 核心摘要：研究人形全身运动的持续学习，提出 Similarity-guided LoRA-PNN，并在六类序列技能与真机上验证。

# 持续人形机器人运动学习

::: v-pre
> 原文链接：[Continual Humanoid Motion Learning](https://arxiv.org/abs/2610.04231)（arXiv 2610.04231 · 发布日期：2026-10-03）

**Continual Humanoid Motion Learning**

研究人形全身运动的持续学习，提出 Similarity-guided LoRA-PNN，并在六类序列技能与真机上验证。

## 核心要点

- 现有全身控制器多离线训练后冻结，学习新技能易侵蚀已掌握技能。
- 提出 Similarity-guided LoRA-PNN，用 PNN 避免遗忘，并以低秩适配复用知识。
- 用动态时间规整与最优传输构成两级相似度，决定依托旧技能和新增容量。
- 六类序列技能上取得最佳前向迁移和最高平均准确率，代码已公开。

**验证范围与局限**：验证范围包括六类序列学习技能、sim-to-sim transfer，并部署于一台物理 Unitree G1。

## 英文摘要 (Abstract)

Humanoid whole-body controllers can now track a diverse set of dynamic motions, but they are typically trained offline and then frozen, so teaching such a controller a new skill tends to erode the skills it already mastered. We study continual learning for humanoid whole-body motion, where a single controller must acquire skills from a sequential task stream without revisiting past data. We introduce Similarity-guided LoRA-PNN, a progressive neural network (PNN) policy that prevents catastrophic forgetting by construction while reusing knowledge across skills through lightweight low-rank adaptation. A two-level motion-similarity measure, built from dynamic time warping aggregated by optimal transport, decides which prior skill to build on and how much new capacity to allocate, yielding strong forward transfer and large efficiency gains. Across six sequentially learned skill categories, our similarity-guided LoRA policy attains the best forward transfer (0.125 vs. 0.079) and the highest average accuracy among all methods, while saving up to 94.5% of trainable parameters and 40.8% of training time. The resulting controller reaches 96.13% sim-to-sim transfer and is deployed on a physical Unitree G1. Our code is available at https://anonymous.4open.science/r/continual-humanoid-learning-35D3.

作者：Zhewen He, Hao Huang, Geeta Chandra Raju Bethala, Chong Yu, Tao Chen, Anthony P. Tzes, Yi Fang
:::

## 原始参考来源 (Sources)

- [arXiv 2610.04231：Continual Humanoid Motion Learning](https://arxiv.org/abs/2610.04231)