# Words into Action: Learning Diverse Humanoid Robot Behaviors using Language Guided Iterative Motion Refinement

arXiv: https://arxiv.org/abs/2310.06226 · HTML: https://mlsanity.com/paper/2310.06226/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 10 Oct 2023.

## Built on by (most cited first, 8 of 8)

- [Agent AI: Surveying the Horizons of Multimodal Interaction](https://mlsanity.com/paper/2401.03568/) (2024)
- [Expressive Whole-Body Control for Humanoid Robots](https://mlsanity.com/paper/2402.16796/) (2024)
- [Visual Whole-Body Control for Legged Loco-Manipulation](https://mlsanity.com/paper/2403.16967/) (2024)
- [AnySkill: Learning Open-Vocabulary Physical Skill for Interactive Agents](https://mlsanity.com/paper/2403.12835/) (2024)
- [An Interactive Agent Foundation Model](https://mlsanity.com/paper/2402.05929/) (2024)
- [Harmon: Whole-Body Motion Generation of Humanoid Robots from Language Descriptions](https://mlsanity.com/paper/2410.12773/) (2024)
- [Deep Reinforcement Learning for Bipedal Locomotion: A Brief Survey](https://mlsanity.com/paper/2404.17070/) (2024)
- [Position Paper: Agent AI Towards a Holistic Intelligence](https://mlsanity.com/paper/2403.00833/) (2024)

## Abstract (arXiv)

Humanoid robots are well suited for human habitats due to their morphological similarity, but developing controllers for them is a challenging task that involves multiple sub-problems, such as control, planning and perception. In this paper, we introduce a method to simplify controller design by enabling users to train and fine-tune robot control policies using natural language commands. We first learn a neural network policy that generates behaviors given a natural language command, such as "walk forward", by combining Large Language Models (LLMs), motion retargeting, and motion imitation. Based on the synthesized motion, we iteratively fine-tune by updating the text prompt and querying LLMs to find the best checkpoint associated with the closest motion in history. We validate our approach using a simulated Digit humanoid robot and demonstrate learning of diverse motions, such as walking, hopping, and kicking, without the burden of complex reward engineering. In addition, we show that our iterative refinement enables us to learn 3x times faster than a naive formulation that learns from scratch.