This repository serves as a curated collection of resources for understanding and applying diffusion models in robotics. It targets researchers and engineers interested in leveraging generative AI for robotic tasks, offering a structured overview of key papers, blogs, and code implementations. The primary benefit is a centralized, categorized knowledge base that accelerates learning and project initiation in this rapidly evolving field.

How It Works

The repository categorizes diffusion model applications in robotics into sub-fields like imitation learning, video generation, reinforcement learning, and task planning. It provides links to foundational diffusion model papers and tutorials, explaining the core concepts from both denoising diffusion probabilistic models (DDPM) and score-based generative models, unified by stochastic differential equations (SDEs). This structured approach allows users to quickly identify relevant research and understand the underlying principles.

Quick Start & Requirements

• Installation: No direct installation is required as this is a literature repository.
• Prerequisites: Access to the internet to view linked papers, blogs, and code repositories.
• Resources: Standard web browsing capabilities.
• Links:
  • Learning about Diffusion models
  • Diffusion in Robotics
  • Code Implementations

Highlighted Details

• Comprehensive categorization of diffusion model applications across various robotics sub-domains.
• Links to seminal papers and influential blog posts for foundational understanding.
• Curated list of relevant code repositories for practical implementation.
• Overview of diffusion model history and theoretical perspectives (DDPM, score-based models, SDEs).

Maintenance & Community

• The repository is maintained by mbreuss.
• No specific community links (Discord/Slack) or roadmap are provided in the README.

Licensing & Compatibility

• The repository itself does not have a specified license. The linked code repositories will have their own licenses.

Limitations & Caveats

The literature is rapidly expanding, meaning the listed papers may not represent the absolute latest advancements. The repository focuses on providing a starting point rather than an exhaustive, real-time database.