Vincent2025hello/u0_pi05

U0 π0.5 — Underwater Robot VLA Model (Baseline)

Model ID: Vincent2025hello/u0_pi05
Base Model: π0.5 (gs://openpi-assets/checkpoints/pi05_base)
License: Gemma (see openpi LICENSE_GEMMA)
Paper: USIM and U0: A Vision-Language-Action Dataset and Model for General Underwater Robots

---

Model Description

This model is the π0.5 baseline of the paper — a Vision-Language-Action (VLA) policy fine-tuned from the official π0.5 base weights on the USIM underwater dataset for the U0 underwater robot (based on BlueROV2). It takes camera observations and multi-sensor state inputs, and outputs action chunks for autonomous underwater tasks.

Fine-Tuning Details

Item	Value
Base Model	π0.5 base checkpoint
Fine-Tuning Method	Frozen PaliGemma LLM; trainable vision tower & action expert; absolute actions
Training Steps	22000 (2 epochs), batch size 64
Checkpoint	Final checkpoint saved at step 21999
Train Config	pi05_u0bot
Dataset	Vincent2025hello/usim

Input / Output

Inputs

• Images: ego-view + wrist-view images
• State (29-dim):
  • joint_pos (5): joint positions
  • pwm (8): thruster PWM values
  • joint_v (5): joint velocities
  • dvl_v (3): DVL velocity
  • imu_av (3): IMU angular velocity
  • imu_la (3): IMU linear acceleration
  • pressure (1): depth pressure
  • dvl_h (1): DVL altitude
• Language: natural language task description

Outputs

• Action (13-dim × 16 steps):
  • joint_pos (5): target joint positions
  • pwm (8): target thruster PWM values

Download Model

pip install huggingface_hub
hf download Vincent2025hello/u0_pi05 --local-dir ./u0_pi05

Usage

The complete fine-tuning, evaluation, and inference framework (including a GR00T-compatible HTTP inference service) is available at: https://github.com/VincentGu2000/u0-openpi

Note: serving a policy loads norm_stats.json from the USIM dataset directory — download the dataset and run the compute-norm-stats step from the repository README first.

Citation

@misc{gu2025usimu0visionlanguageactiondataset,
      title={USIM and U0: A Vision-Language-Action Dataset and Model for General Underwater Robots}, 
      author={Junwen Gu and Zhiheng Wu and Pengxuan Si and Shuang Qiu and Yukai Feng and Luoyang Sun and Laien Luo and Lianyi Yu and Jian Wang and Zhengxing Wu},
      year={2025},
      eprint={2510.07869},
      archivePrefix={arXiv},
      primaryClass={cs.RO},
      url={https://arxiv.org/abs/2510.07869}, 
}

Acknowledgments

This model is fine-tuned from π0.5. We thank the Physical Intelligence team for open-sourcing the π0 / π0.5 models and the openpi framework.