Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Awesome collection of resources and papers on Diffusion Models for Robotic Manipulation.
| Date | Stars |
|---|---|
| 2026-07-31 | 817 |
| 2026-08-01 | 817 |
| 2026-08-06 | 817 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
<div align="center">
# A Survey on Diffusion Policy for Robotic Manipulation
</div>
<div align="center">
Mingchen Song,
<a target="_blank" href="https://homepage.hit.edu.cn/dengxiang">Xiang Deng</a>,
Zhiling Zhou,
Jie Wei,
<a target="_blank" href="https://ieeexplore.ieee.org/author/37087008154">Weili Guan</a>,
<a target="_blank" href="https://scholar.google.com/citations?hl=en&user=yywVMhUAAAAJ">Liqiang Nie</a>
<br>
</div>
<br>
🔥 Since 2022, research on diffusion policies for robotic manipulation has demonstrated consistently superior performance compared to traditional methodologies across diverse tasks. Despite the rapid growth and promising results in this field, there remains a notable absence of comprehensive survey literature that systematically analyzes and synthesizes developments in this evolving research field.
<div style="text-align: center;">
<img src="./timeline.jpg" alt="image info">
</div>
<table>
<tr>
<td>
<img src="./number_of_paper_page.jpg" alt="Diffusion Policy Papers Over Time" />
</td>
<td>
<img src="./model.jpg" alt="Comparing Architectures of Different Diffusion Policy Models." />
</td>
</tr>
</table>
📍 We present **the first comprehensive analysis of diffusion policies for robotic manipulation**. Specifically, we systematically analyze existing methods from three perspectives: data representation, model architecture, and diffusion strategy. For example, from the model architecture perspective, we categorize existing diffusion policy methods into three types: (a) Large Language Model Based Diffusion Policy, (b) Small Size CNN or Transformer Based Diffusion Policy, and (c) VAE / VQ-VAE Based Diffusion Policy.
🚀 For a deeper dive, please check out our [survey paper](https://doi.org/10.36227/techrxiv.174378343.39356214/v1): **A Survey on Diffusion Policy for Robotic Manipulation: Taxonomy, Analysis, and Future Directions**
## 📑 Table of Contents
- 🤖[A Survey on Diffusion Policy for Robotic Manipulation](#a-survey-on-diffusion-policy-for-robotic-manipulation)
- 📑[Table of Contents](#-table-of-contents)
- 📖[Papers](#-papers)
- 📊[Data Representation](#-data-representation)
- [2D Representations](#2d-representations)
- [3D Representations](#3d-representations)
- [Heterogeneous Data](#heterogeneous-data)
- 🧠[Model Architecture](#-model-architecture)
- [Large Language Model Based Diffusion Policy](#large-language-model-based-diffusion-policy)
- [Small Size CNN or Transformer Model Based Diffusion Policy](#small-size-cnn-or-transformer-model-based-diffusion-policy)
- [VAE / VQ-VAE Based Diffusion Policy](#vae--vq-vae-based-diffusion-policy)
- 🌊[Diffusion Strategy](#-diffusion-strategy)
- [Incorporating Reinforcement Learning](#incorporating-reinforcement-learning)
- [Combined with Equivariance](#combined-with-equivariance)
- [Accelerated Sampling or Denoising Strategies](#accelerated-sampling-or-denoising-strategies)
- [Employing Classifier (free) Guidance](#employing-classifier-free-guidance)
- [Integration with Self-Supervised Learning](#integration-with-self-supervised-learning)
- 🦾[Simulation Platforms & Real-World Robots](#-simulation-platforms--real-world-robots)
- 📜[Citation](#-citation)
## 📖 Papers
### 📊 Data Representation
#### 2D Representations
- **Diffusion Policy Policy Optimization**, ICLR 2025. [[paper](https://arxiv.org/abs/2409.00588)] [[code](https://github.com/irom-princeton/dppo)] [[website](https://diffusion-ppo.github.io/)]
- **Human2Robot: Learning Robot Actions from Paired Human-Robot Videos**, arXiv 2025. [[paper](https://arxiv.org/abs/2502.16587)]
- **Task-Agnostic Pre-training and Task-Guided Fine-tuning**, ICML 2025. [[paper](https://arxiv.org/abs/2409.19949)]
- **Latent Action Pretraining from Videos**, ICLR 2025. [[paper](https://arxiv.org/abs/2410.11758)] [[code](https://github.com/LatentActionPretraExcerpt of 58,920 characters
Read on GitHub
40
9
4
Would you bet a product on this? Bounded 0–100 and slow moving.
matched fp:abbb62609abec178, name:robotics, desc:manipulation