Awesome

M2DGR-plus: Extension and update of M2DGR, a novel Multi-modal and Multi-scenario SLAM Dataset for Ground Robots (ICRA2022 & ICRA2024)

First Author: Jie Yin 殷杰 📝 [Paper] ➡️ [Algorithm] 🎯 [M2DGR Dataset] ⭐️ [Presentation Video]

<div align=center> <img src="./fig/car2.jpg" width="800px"> </div> <p align="center">Figure 1. Acquisition Platform and Diverse Scenarios.</p>

News & Updates

🔥2024/10/11: Introducing M2DGR-benchmark, benchmarking newest SOTA LiDAR-visual SLAM alrogithms on both M2DGR and M2DGR-plus!
2024/07/15: Introducing a list of LiDAR-Visual SLAM systems at awesome-LiDAR-Visual-SLAM, wheel-based SLAM systems at awesome-wheel-slam, and Isaac Sim resources at awesome-isaac-sim (keep updating)

This dataset is based on M2DGR. And the algorithm code is Ground-Fusion. The preprint version of this paper is arxiv.

1.LICENSE

This work is licensed under MIT license. International License and is provided for academic purpose. If you are interested in our project for commercial purposes, please contact us on 1195391308@qq.com for further communication.

If you use this work in an academic work, please cite:

@article{yin2021m2dgr,
  title={M2dgr: A multi-sensor and multi-scenario slam dataset for ground robots},
  author={Yin, Jie and Li, Ang and Li, Tao and Yu, Wenxian and Zou, Danping},
  journal={IEEE Robotics and Automation Letters},
  volume={7},
  number={2},
  pages={2266--2273},
  year={2021},
  publisher={IEEE}
}

@article{yin2024ground,
  title={Ground-Fusion: A Low-cost Ground SLAM System Robust to Corner Cases},
  author={Yin, Jie and Li, Ang and Xi, Wei and Yu, Wenxian and Zou, Danping},
  journal={arXiv preprint arXiv:2402.14308},
  year={2024}
}

2.SENSOR SETUP

The calibration results are here. All the sensors and track devices and their most important parameters are listed as below:

LIDAR Robosense 16, 360 Horizontal Field of View (FOV),-30 to +10 vertical FOV,5Hz,Max Range 200 m,Range Resolution 3 cm, Horizontal Angular Resolution 0.2°.
GNSS Ublox F9p, GPS/BeiDou/Glonass/Galileo, 1Hz
V-I Sensor,Realsense d435i,RGB/Depth 640*480,69H-FOV,42.5V-FOV,15Hz;IMU 6-axix, 200Hz
IMU,wheeltec,9-axis,100Hz;
GNSS-IMU Xsens Mti 680G. GNSS-RTK,localization precision 2cm,100Hz;IMU 9-axis,100 Hz;
Motion-capture System Vicon Vero 2.2, localization accuracy 1mm, 50 Hz;

The rostopics of our rosbag sequences are listed as follows:

3D LIDAR: /rslidar_points
2D LIDAR: /scan
Odom: /odom
GNSS Ublox F9p:
/ublox_driver/ephem ,

/ublox_driver/glo_ephem ,

/ublox_driver/range_meas ,

/ublox_driver/receiver_lla ,

/ublox_driver/receiver_pvt

V-I Sensor:
/camera/color/image_raw,
/camera/imu
IMU: /imu

3.DATASET SEQUENCES

Sequence Name	Collection Date	Total Size	Duration	Features	Rosbag
Anomaly	2023-8	1.5g	57s	wheel anomaly	Rosbag
Switch	2023-8	9.5g	292s	indoor-outdoor switch	Rosbag
Tree	2023-8	3.7g	160s	Dense tree leave cover	Rosbag
Bridge_01	2022-11	2.4g	75s	Bridge, Zigzag	Rosbag
Bridge_02	2022-11	16.0g	501s	Bridge, Long-term,Straight line	Rosbag
Street_01	2022-11	1.7g	58s	Street, Straight line	Rosbag
Street_02	2022-11	3.9g	126s	Bridge, Sharp turn	Rosbag
Parking_01	2022-11	3.3g	105s	Parking lot, Side moving	Rosbag
Parking_02	2022-11	5.4g	149s	Parking lot, Rectangle loop	Rosbag
Building_01	2022-11	3.7g	120s	Building, Far features	Rosbag
Building_02	2022-11	3.4g	110s	Building, Far features	Rosbag

</div>

4. EXPERIMENTAL RESULTS

We test methods with diverse senser settings to validate our benchmark dataset. Results shown that our dataset is a valid and effective testfield for localization methods.

And in some cases, our Ground-Fusion achieves comparable performance to Lidar SLAM!

<div align=center> <img src="./fig/resultf.png" width="800px"> </div> <p align="center">Figure 2. The ATE RMSE (m) result on some sequences.</p> <div align=center> <img src="./fig/result.png" width="800px"> </div> <p align="center">Figure 3. The visualized trajectory.</p>

5. Configuration Files

We provide configuration files for several cutting-edge baseline methods, including VINS-RGBD,TartanVO,VINS-Mono and VIW-Fusion and GVINS.