FusionPortable

A Multi-Sensor Campus-Scene Dataset for Evaluation of Localization and Mapping Accuracy on Diverse Platforms

Introduction

We consider that a desirable dataset should fulfill the following four requirements:

  1. Various sensors (LiDARs, cameras, IMU, etc.).
  2. Various robotic platforms with diverse motion patterns.
  3. Sequences cover from room-scale (meter-level) to large-scale (kilometer-level).
  4. Benchmarking for different tasks.

We are motivated to propose the FusionPortable dataset, which is initially intended to support odometry, localization, mapping, and some perception tasks. We advance a self-contained, portable, and versatile multi-sensor suite. We construct a dataset that covers a variety of environments on the campus by exploiting multiple robot platforms for data collection. We also provide ground truth for the decouple localization and mapping performance evaluation.

Data Collection Platforms

Sensors

Robots

Examples of Data Collection

Environment Platform Preview
Garden Handheld
Motion Capture Room Quadrupled Robot

Download

Data Organization


FusionPortable/
├── calibration_files/               // Intrinsics & extrinsics of sensors
  └── 20220209_calib/
    └── <sensor_name>.yaml           // e.g., ouster00.yaml, frame_cam00.yaml 
├── groundtruth/
  └── map/                           // Ground-truth maps
    └── <date_env>/
      ├── scan/
        └── <scan_id>.pcd            // Individual scan
      ├── merged_scan.pcd            // Merged scan (resolution: 1cm)
      └── transformation.yaml        // Transformation of each scan
  └── traj/                          // Ground-truth trajectories 
    └── <date_env>.txt               // e.g., 20220215_canteen_night.txt
└── sensor_data/
  └── <platform>/                    // Platforms, e.g., handheld
    └── <date_env>                   // e.g., 20220215_canteen_night
      ├── <date_env>.bag             
      ├── <date_env>.bag.7z  
      ├── data/
      └── data_ref_kitti/

Note:

1. <date_env>.bag is the raw rosbag.
2. data/ stores indivisual sensor data files with timestamps from timestamps.txt.
3. data_ref_kitti/ follows the KITTI format to store sensor data files from data/.

Download

Please click these below links to download:

1. sensor_data - pwd: fusionportable
2. ground-truth trajectories and maps - pwd: fusionportable
3. calibration_files - pwd: fusionportable

Note: Extract the ROS bag from .7z files in the terminal: 7z e <filename>.7z

Sequences

Type Platform Picture Sequence Preview
Calibration Handheld Motion Capture Room 20220209_StaticTarget_SmallCheckerBoard_9X12_30mm
Calibration Handheld Motion Capture Room 20220215_DynamicTarget_BigCheckerBoard_7X10_68mm
Calibration Handheld Motion Capture Room 20220209_Static_IMUs_3h20mins
Handheld Canteen 20220216_canteen_night preview
Handheld Canteen 20220216_canteen_day preview
Handheld Garden 20220215_garden_night preview
Handheld Canteen 20220216_garden_day preview
Handheld Canteen 20220216_corridor_day preview
Handheld Canteen 20220216_escalator_day preview
Handheld Buliding 20220225_building_day preview
Handheld Motion Capture Room 20220216_MCR_slow preview
Handheld Motion Capture Room 20220216_MCR_normal preview
Handheld Motion Capture Room 20220216_MCR_fast preview
Quadruped Robot MCR_slow_00 20220219_MCR_slow_00 preview
Quadruped Robot MCR_slow_00 20220219_MCR_slow_01 preview
Quadruped Robot MCR_slow_00 20220219_MCR_normal_00 preview
Quadruped Robot MCR_slow_00 20220219_MCR_normal_01 preview
Quadruped Robot MCR_slow_00 20220219_MCR_fast_00 preview
Quadruped Robot MCR_slow_00 20220219_MCR_fast_01 preview
Apollo Vehicle Campus Road 20220226_campus_road preview

Some High-Resolution GT Maps

Environment Platform
Garden GT Map_Garden
Escalator GT Map_Escalator
Building GT Map_Building

Tools

The dataset tool can be used by clicking the button below

Evaluation

Evalaution of Trajectories

Localization Accuracy

Issues

If you have any issues with the theme, please report them on the repository:

Publications

  1. FusionPortable: A Multi-Sensor Campus-Scene Dataset for Evaluation of Localization and Mapping Accuracy on Diverse Platforms
    Jianhao Jiao*, Hexiang Wei*, Tianshuai Hu*, Xiangcheng Hu*, Yilong Zhu, Zhijian He, Jin Wu, Jingwen Yu, Xupeng Xie, Huaiyang Huang, Ruoyu Geng, Lujia Wang, Ming Liu
    Presented at IROS 2022
    [paper] [bibtex]