The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
Construction-Site Multimodal Robotics ROS 2 Recordings
Construction-Site Multimodal Robotics ROS 2 Recordings
This repository contains the raw ROS 2 MCAP recordings collected from construction-site environments for the accompanying study on failure-mode-aware traversability mapping for autonomous mobile robots (AMRs).
The recordings preserve the multimodal sensor streams and their temporal relationships during real robot operation. All four recording sessions are closed-loop robot trajectories, providing continuous sensor data while the robot operated and revisited regions of the construction sites.
Dataset Structure
construction-traversability-rosbags/
├── site1/
│ ├── site1_session01/
│ │ ├── metadata.yaml
│ │ └── site1_session01.mcap
│ └── site1_session02/
│ ├── metadata.yaml
│ └── site1_session02.mcap
└── site2/
├── site2_session01/
│ ├── metadata.yaml
│ └── site2_session01.mcap
└── site2_session02/
├── metadata.yaml
└── site2_session02.mcap
Each session contains an MCAP recording and its corresponding ROS 2 metadata.yaml file.
Sensor Data
The recordings contain multimodal ROS 2 data collected from the AMR, including:
- OAK-D RGB images
- OAK-D depth images
- Livox 3D LiDAR point clouds
- IMU measurements
- Robot odometry
- TF and static TF information
- Associated timestamps
The exact topics, message counts, recording duration, and other metadata can be inspected using standard ROS 2 bag tools.
Closed-Loop Sessions
The four recordings correspond to two construction sites, with two closed-loop sessions recorded at each site:
- Site 1:
site1_session01,site1_session02 - Site 2:
site2_session01,site2_session02
The closed-loop nature of the recordings preserves continuous multimodal observations over robot trajectories rather than isolated sensor captures. This makes the recordings suitable for research involving temporal sensor fusion, mapping, localization, loop-closure analysis, and autonomous navigation.
Related Annotated Dataset and Trained Model
The curated annotated RGB dataset, camera calibration files, dataset configuration, semantic class taxonomy, and the fine-tuned YOLOv26-m-seg model used in the accompanying experiments are provided in the main dataset repository:
Construction-Traversability Dataset:
https://huggingface.co/datasets/manojkarnekar/construction-traversability-dataset
The main dataset contains 506 annotated RGB frames spanning 28 construction-site semantic classes. It also provides the calibration information required for RGB-LiDAR projection and multimodal fusion.
The trained semantic segmentation checkpoint is released to facilitate reproduction of the semantic perception and RGB-LiDAR fusion experiments.
Data Format
The recordings are provided in ROS 2 MCAP format.
Each MCAP recording is accompanied by its metadata.yaml file. The bags can be inspected using:
ros2 bag info <path_to_bag_directory>
and played using:
ros2 bag play <path_to_bag_directory>
A ROS 2 installation with MCAP storage support is required.
Intended Use
The recordings are intended to support research in:
- Construction-site robotic perception
- LiDAR-camera sensor fusion
- Semantic point-cloud generation
- Traversability mapping
- Local costmap generation
- Autonomous navigation
- Localization and mapping
- Loop-closure analysis
- Multimodal and temporal perception
The raw recordings can also be used to create additional annotations or alternative train/test splits beyond those used in the accompanying work.
Privacy
For privacy preservation, persons, textual information, and vehicle number plates visible in the collected imagery were blurred during dataset preparation.
License
This dataset is released under the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license.
The recordings may be used, shared, and adapted for non-commercial purposes with appropriate attribution. Commercial use is not permitted without separate permission from the dataset authors.
See the accompanying LICENSE file for the complete license terms.
- Downloads last month
- 46