PredActor

Model family: PredActor is an onboard humanoid diffusion policy that jointly predicts future states and executable actions from proprioception. It supports text-driven motion and guidance toward task objectives. The project reports 50 Hz onboard control on Unitree G1.

Papers & reports (1)

PredActor: Predictive Action Diffusion for Steerable Onboard Humanoid Control

Paper available

arXiv: 2609.24840
Authors & affiliations (13)

Author order: Original order

  1. Lei Ye

    Not provided

  2. Haibo Gao

    Not provided

  3. Yitang Li

    Not provided

  4. Peng Xu

    Not provided

  5. Zetong Jing

    Not provided

  6. Junhan Sun

    Not provided

  7. Fanrong Dong

    Not provided

  8. Ziqi Han

    Not provided

  9. Xue Wang

    Not provided

First published
2026-09-21
Latest paper revision
Not provided
Venue
arXiv

Covered model versions

Paper ↗PDF ↗

How to access and use

Public code resources
Public access
Model files
Public access

Training / fine-tuning code

Access
Public access

Verified on · 2026-11-21

Code license
MIT

Model files

Model files
PredActor · PDP051AccessWeights license
PredActor · PDP067 G1 deploymentAccessWeights license
Release information

SourceVerified on · 2026-11-21

Checks before deployment

Available weights or code do not establish plug-and-play robot compatibility. Check the required configuration, data and interfaces in the source documentation.

Deployment and usage resources

Architecture and inputs / outputs

Model type
Action policy
Action generation
Diffusion
Input
Language / Robot state

Scope and training

Output

Actions

Robot form

Humanoid

Submit a source or correction