This dataset contains the raw ROS bag files for a subset of the AIROA household service robot teleoperation dataset. The data was collected using Toyota's Human Support Robot (HSR) platform for primitive action (PA) household manipulation tasks.

Key Statistics

Total Episodes: 7 task demonstrations
Total Size: ~6.4 GB (compressed rosbag files)
Episode Duration Range: 10.1s - 441s
Recording Format: ROS bag (version 2.0)
Metadata Version: 1.3
Robots Used: 5 unique HSR robots
Operators: 5 operators (anonymized)
Location: Single laboratory environment

Task Distribution

Task Category	Episodes	Bag Size	Duration
Pull Chain Lamp	2	86-90 MB	44-51s
Toast Baking	1	847 MB	142s
Coffee Making	1	1.7 GB	317s
Dishwashing	1	2.5 GB	441s
Towel Manipulation	1	629 MB	104s
Slipper Organization	1	840 MB	125s

Data Structure

Each episode directory contains:

<template_id>/
├── data.bag              # ROS bag file with sensor data
├── meta.json         # Episode metadata (v1.3 schema)

ROS Topics

All rosbag files include the following ROS topics:

Camera Streams

/hsrb/head_rgbd_sensor/rgb/image_rect_color/compressed - Head RGB camera (640×480, compressed)
/hsrb/head_rgbd_sensor/depth_registered/image_rect_raw/compressedDepth - Head depth camera
/hsrb/hand_camera/image_raw/compressed - Hand RGB camera (640×480, compressed)
/hsrb/head_rgbd_sensor/rgb/camera_info - Head camera calibration
/hsrb/hand_camera/camera_info - Hand camera calibration

Robot State

/hsrb/joint_states - Joint positions, velocities, efforts
/hsrb/odom - Robot odometry
/hsrb/base_scan - Laser scan data
/hsrb/wrist_wrench/raw - Force/torque sensor data
/hsrb/servo_states - Servo controller states

Control Commands

/hsrb/arm_trajectory_controller/command - Arm motion commands
/hsrb/command_velocity - Base velocity commands
/hsrb/gripper_controller/command - Gripper commands
/hsrb/head_trajectory_controller/command - Head motion commands
/control_mode - Control mode indicator

Transforms

/tf - Transform tree (coordinate frames)
/tf_static - Static transforms

Message Types

geometry_msgs/Twist                           - Velocity commands
geometry_msgs/WrenchStamped                   - Force/torque data
nav_msgs/Odometry                             - Odometry data
sensor_msgs/CameraInfo                        - Camera calibration
sensor_msgs/CompressedImage                   - Compressed RGB images
sensor_msgs/JointState                        - Joint states
sensor_msgs/LaserScan                         - Laser scan data
std_msgs/String                               - Control mode
tf2_msgs/TFMessage                            - Transforms
tmc_control_msgs/GripperApplyEffortActionGoal - Gripper goals
tmc_control_msgs/ServoState                   - Servo states
trajectory_msgs/JointTrajectory               - Trajectory commands

Metadata Schema (v1.3)

Each episode includes structured metadata with:

UUID: Unique episode identifier
Context: Robot ID, location, software versions
Run: Task instructions, temporal segments, success indicators
Files: Associated data files (rosbag)

Usage

Converting to LeRobot Format with Rebake

This raw dataset is designed to be converted to the standardized LeRobot format using the rebake utility.

Installation

See the rebake README for installation instructions.

Basic Conversion

# Convert the airoa-moma-raw dataset to LeRobot format
uv run -m hsr_data_converter.rosbag2lerobot.main \
    --raw_dir /path/to/airoa-moma-raw \
    --out_dir /path/to/output \
    --fps 10 \
    --robot_type hsr \
    --conversion_type aggregate

This will process all rosbag files and their metadata, generating a LeRobot-compatible dataset with:

Synchronized video frames from head and hand cameras
Robot state observations (joint positions, velocities, etc.)
Action trajectories
Episode metadata and task annotations

For advanced usage, filtering, and visualization options, see the rebake documentation.

Direct ROS Usage

For direct rosbag inspection: rosbag info <template_id>/data.bag

For metadata parsing and validation, see the airoa-metadata repository.

Hardware Configuration

Robots

Platform: Toyota Human Support Robot (HSR)
Sensors:
- Head RGBD camera (640×480 @ 30fps)
- Hand RGB camera (640×480 @ 30fps)
- Base laser scanner
- 6-axis wrist force/torque sensor
- Joint encoders

Environment

Location: Single laboratory environment
Tasks: Household manipulation primitives
Objects: Common household items (towels, dishes, coffee maker, etc.)

Data Collection

Interface: HSR Leader Teleop (human teleoperation)
Control: Whole-body teleoperation with leader robot
Recording: Continuous recording during task execution
Segmentation: Automatic temporal segmentation based on instructions

Related Datasets

AIROA-MOMA: Processed dataset with extracted frames and annotations https://huggingface.co/datasets/airoa-org/airoa-moma

Citation

If you use this dataset in your research, please cite:

@article{airoa-moma-2025,
  author  = {Ryosuke Takanami, Petr Khrapchenkov, Shu Morikuni, Jumpei Arima, Yuta Takaba, Shunsuke Maeda, Takuya Okubo, Genki Sano, Satoshi Sekioka, Aoi Kadoya, Motonari Kambara, Naoya Nishiura, Haruto Suzuki, Takanori Yoshimoto, Koya Sakamoto, Shinnosuke Ono, Yo Ko, Daichi Yashima, Aoi Horo, Tomohiro Motoda, Kensuke Chiyoma, Hiroshi Ito, Koki Fukuda, Akihito Goto, Kazumi Morinaga, Yuya Ikeda, Riko Kawada, Masaki Yoshikawa, Norio Kosuge, Yuki Noguchi, Kei Ota, Tatsuya Matsushima, Yusuke Iwasawa, Yutaka Matsuo, Tetsuya Ogata},
  title   = {AIRoA MoMa Dataset: A Large-Scale Hierarchical Dataset for Mobile Manipulation},
  journal = {arXiv preprint},
  year    = {2025}
}

Contact

For questions or issues regarding this dataset, please contact the AIROA team or open an issue in the repository.

Dataset Version: 1.0 Last Updated: October 2025 Schema Version: 1.3

Downloads last month: 9