TidyBot is a dataset developed by Princeton University for household cleaning tasks. It contains 570 episodes of a PR2 robot performing cleaning tasks like sweeping and mopping, including RGB images, depth data, and robot joint states. The dataset supports research in hierarchical imitation learning and multi-stage task planning, with natural language instructions and visual goals. It is accompanied by a detailed benchmark and evaluation framework, making it suitable for studying long-horizon manipulation and real-world industrial automation. While the dataset's license is not explicitly stated, it is primarily intended for academic use and emphasizes the integration of language and vision for task execution.
The robot puts each object into the appropriate receptacle based on user preferences
| Field | Value |
|---|---|
| Action Space | Our dataset specifies a target receptacle for each object |
| Depth Cams | 0 |
| Gripper | Robotiq 2F-85 |
| Has Camera Calibration | False |
| Has Proprioception | False |
| Has Suboptimal | False |
| Language Annotations | Our dataset specifies object placements in text form |
| Rgb Cams | 0 |
| Robot Morphology | Mobile Manipulator |
| Scene Type | Kitchen (also toy kitchen), Other Household environments, living room, bedroom, kitchen, pantry room |
| Wrist Cams | 0 |