Reactive Diffusion Policy
- AI model
- Humanoids and robot learning
A slow-fast visual-tactile policy learning system for contact-rich robot manipulation, built on diffusion policies and tested on bimanual real robots.
Source on GitHub: github.com/xiaoxiaoxh/reactive_diffusion_policy (opens in a new tab). Facts read at commit 824c5e8 (opens in a new tab).
Licenses
- Code
- Apache-2.0 (opens in a new tab)
Read by Robotheca from LICENSE (opens in a new tab) (Apache-2.0) at commit 824c5e8 (opens in a new tab), on 16 Oct 2026.
- Weights
- MIT (opens in a new tab)
Read on huggingface.co/WendiChen/reactive_diffusion_policy_model (opens in a new tab).
- Commercial use
- Yes
What it does
- Learns robot manipulation policies that combine visual and tactile sensing for contact-rich tasks.
- Includes data collection tooling built around a Meta Quest 3 teleoperation app and ROS2 services.
- Provides post-processing that converts recorded data into Zarr format for training.
- Offers training scripts for Diffusion Policy and Reactive Diffusion Policy with configurable control frequencies.
- Supports inference by running teleoperation, camera launcher and evaluation services in separate terminals.
- Extends to Franka Research 3 arms and single-arm teleoperation, with a guide for customized sensors and robots.
Written by Robotheca from the project's README (opens in a new tab) at commit 824c5e8.
Getting started
- Difficulty
- Advanced
- Hardware it needs, as the project states it
- A workstation running Ubuntu 22.04 with a GPU (e.g., NVIDIA RTX 3090), plus a Meta Quest 3 headset, 2 robot arms (Flexiv Rizon 4 by default, or Franka Research 3), 1-3 RealSense cameras, and optional GelSight Mini tactile sensors; a high-performance CPU (e.g., Core i9-13900K) is required for 24 FPS
- How to start
- The project's website (opens in a new tab)
Maintenance
- Status
- Maintained: a commit in the last 12 months
- Newest commit
- 12 Apr 2026 on main
- Latest release
- None published on GitHub
Checked on GitHub on 16 Oct 2026.
Links
Robots that can run it
No robot on Robotheca is linked to it yet. Links come from a robot's robot.yaml, from its dependency files read as text, or from Robotheca's editors.
Robotheca lists AI models as a reference: nothing is hosted or run here, and every fact links to where it was read.