metadata
tags:
- robot-learning
- ros2
- lerobot
- robot-manipulation
library_name: datasets
Robot Arm Dataset (HuggingFace Viewable)
This dataset contains robot manipulation data extracted from a ROS2 bag file, processed into a HuggingFace-compatible format for easy viewing and analysis.
Dataset Contents
The dataset includes various sensor and control data from a robot arm:
- Joint States: Position, velocity, and effort of robot joints
- TF Transforms: Robot's kinematic state
- Robot Description: URDF and SRDF data
- Logging Information: Rosout messages
- Servo Node Status: Status information from the robot's servo nodes
- Controller Trajectory: Joint trajectory commands and states
Data Format
This dataset is in HuggingFace datasets format with:
- 4,245 samples of robot manipulation data
- Train split with all data
- Compatible with HuggingFace dataset viewer
- Each sample contains observations and actions
Usage
You can load this dataset using HuggingFace datasets:
Data Structure
Each sample contains:
- : Joint positions (14 values)
- : Joint actions (14 values)
- : Various sensor data
- : Time information
- : Frame number
- : Episode number
Visualization
This dataset is compatible with HuggingFace's dataset viewer for interactive exploration.
License
[Specify your dataset license here, e.g., MIT, Apache 2.0, etc.]