| --- |
| tags: |
| - Embodied-AI |
| - Robotics |
| - Offline-to-Online-RL |
| - Robot-Dataset |
| - VLA-Model |
| - Vision-Language-Action |
| - Robot-Learning |
| - Imitation-Learning |
| - Real-Robot-Data |
| --- |
| |
| # Robo-ValueRL Dataset |
|
|
| [[Project Page](https://gewu-lab.github.io/robo-valuerl/)] [[GitHub](https://github.com/Open-X-Humanoid/Robo-ValueRL)] [[Model](https://huggingface.co/X-Humanoid/Robo-ValueRL)] [[Paper](#)] |
|
|
| This repository contains the **dataset** for **Robo-ValueRL: Reliable Value Estimation for Offline-to-Online Reinforcement Learning**. |
|
|
| The Robo-ValueRL dataset provides heterogeneous real-robot experience for studying reliable value estimation, value-guided offline policy pretraining, and online residual adaptation. |
|
|
| ## Dataset Description |
|
|
| The Robo-ValueRL dataset contains real-robot trajectories collected on two long-horizon manipulation tasks: |
|
|
| - **Chip Insertion**: millimeter-level precision manipulation requiring PCB grasping, pose adjustment, chip grasping, and insertion. |
| - **Block Disassembly**: generalizable object disassembly requiring robust grasping, separation, and classification behaviors. |
|
|
| The dataset includes: |
|
|
| - **240 hours** of offline demonstrations |
| - **3,000+** online rollout trajectories |
| - Multi-view robot observations |
| - Language task instructions |
| - Robot states and action chunks |
| - Mixed-quality trajectories, including successful demonstrations, corrections, suboptimal behaviors, and failure cases |
| - Value-derived action-quality labels / conditions for policy learning |
| - Online rollout segments for residual adaptation |
|
|
| ## Associated Model |
|
|
| The dataset is released together with the Robo-ValueRL model suite: |
|
|
| [[Robo-ValueRL Model](https://huggingface.co/X-Humanoid/Robo-ValueRL)] |
|
|
| The associated models include a history-conditioned value estimator, a quality-conditioned VLA policy, and an online residual adaptation module. |
|
|
| ## Data Usage |
|
|
| The dataset is designed for: |
|
|
| 1. Training and evaluating history-conditioned value estimators. |
| 2. Studying value reliability under heterogeneous robotic data. |
| 3. Deriving action-quality conditions from value differences. |
| 4. Training quality-conditioned VLA policies. |
| 5. Evaluating value-guided offline-to-online reinforcement learning. |
| 6. Training online residual adapters from value-filtered rollout segments. |
|
|
| ## Task Setup |
|
|
| ### Chip Insertion |
|
|
| A precision manipulation task where the robot must grasp a PCB, adjust it to a feasible insertion pose, grasp a chip, and insert it into millimeter-scale clearance. |
|
|
| ### Block Disassembly |
|
|
| A generalizable manipulation task where the robot must grasp, separate, and classify block components under varied configurations. |
|
|
| ## Key Features |
|
|
| * **Heterogeneous Robot Experience**: Includes successful, suboptimal, corrective, and failed trajectories. |
| * **Offline and Online Data**: Supports both offline pretraining and online improvement studies. |
| * **Value-Oriented Labels**: Provides value-derived action-quality conditions for policy learning. |
| * **Real-Robot Evaluation**: Collected from physical robot manipulation tasks rather than simulation-only benchmarks. |
| * **Offline-to-Online Pipeline Support**: Designed to connect value estimation, policy pretraining, and residual adaptation. |
|
|
| ## Highlights |
|
|
| - 240h offline demonstrations |
| - 3,000+ online rollout trajectories |
| - Two real-robot manipulation tasks |
| - Multi-view visual observations |
| - Language-conditioned task instructions |
| - Action-quality labels / conditions derived from reliable value estimation |
|
|
| ## Recommended Use |
|
|
| This dataset can be used to reproduce the Robo-ValueRL pipeline or to study new methods for: |
|
|
| - value estimation in robotic manipulation |
| - data filtering from mixed-quality demonstrations |
| - quality-conditioned VLA policy learning |
| - offline-to-online reinforcement learning |
| - stable online adaptation from real-world rollouts |
|
|
| Please refer to the [GitHub repository](https://github.com/Open-X-Humanoid/Robo-ValueRL) for data loading, preprocessing, and training scripts. |
|
|
| ## Citation |
|
|
| If you use the Robo-ValueRL dataset in your research, please cite our work. Citation will be updated after the arXiv release. |
|
|
| ## License |
|
|
| Please refer to the license file in the [GitHub repository](https://github.com/Open-X-Humanoid/Robo-ValueRL). |
|
|
| ## Contact |
|
|
| For questions, please open an issue on our [GitHub repository](https://github.com/Open-X-Humanoid/Robo-ValueRL). |
