LIO-SAM
No Label
Robot
|SLAM
|Autonomous Driving
|...
许可协议: BSD-3-Clause

Overview

We now describe a series of experiments to qualitatively and quantitatively analyze our proposed framework. The sensor suite used in this paper includes a Velodyne VLP16 lidar, a MicroStrain 3DM-GX5-25 IMU, and a Reach M GPS. For validation, we collected 5 different datasets across various scales, platforms and environments. These datasets are referred to as Rotation, Walking, Campus, Park and Amsterdam, respectively. The details of these datasets are shown in Table I.

Data Collection

The first three datasets, namely Rotation, Walking, and Campus datasets were collected using a custom-built handheld device on the MIT campus. The Park dataset was collected in a park covered by vegetation, using an unmanned ground vehicle (UGV) - the Clearpath Jackal. The last dataset, Amsterdam, was collected by mounting the sensors on a boat and cruising in the canals of Amsterdam.

Data Format

Prepare lidar data

The user needs to prepare the point cloud data in the correct format for cloud deskewing, which is mainly done in "imageProjection.cpp". The two requirements are:

  • Provide point time stamp. LIO-SAM uses IMU data to perform point cloud deskew. Thus, the relative point time in a scan needs to be known. The up-to-date Velodyne ROS driver should output this information directly. Here, we assume the point time channel is called "time." The definition of the point type is located at the top of the "imageProjection.cpp." "deskewPoint()" function] utilizes this relative time to obtain the transformation of this point relative to the beginning of the scan. When the lidar rotates at 10Hz, the timestamp of a point should vary between 0 and 0.1 seconds. If you are using other lidar sensors, you may need to change the name of this time channel and make sure that it is the relative time in a scan.
  • Provide point ring number. LIO-SAM uses this information to organize the point correctly in a matrix. The ring number indicates which channel of the sensor that this point belongs to. The definition of the point type is located at the top of "imageProjection.cpp." The up-to-date Velodyne ROS driver should output this information directly. Again, if you are using other lidar sensors, you may need to rename this information. Note that only mechanical lidars are supported by the package currently.

Prepare IMU data

  • IMU requirement. Like the original LOAM implementation, LIO-SAM only works with a 9-axis IMU, which gives roll, pitch, and yaw estimation. The roll and pitch estimation is mainly used to initialize the system at the correct attitude. The yaw estimation initializes the system at the right heading when using GPS data. Theoretically, an initialization procedure like VINS-Mono will enable LIO-SAM to work with a 6-axis IMU. The performance of the system largely depends on the quality of the IMU measurements. The higher the IMU data rate, the better the system accuracy. We use Microstrain 3DM-GX5-25, which outputs data at 500Hz. We recommend using an IMU that gives at least a 200Hz output rate. Note that the internal IMU of Ouster lidar is an 6-axis IMU.
  • IMU alignment. LIO-SAM transforms IMU raw data from the IMU frame to the Lidar frame, which follows the ROS REP-105 convention (x - forward, y - left, z - upward). To make the system function properly, the correct extrinsic transformation needs to be provided in "params.yaml" file. The reason why there are two extrinsics is that my IMU (Microstrain 3DM-GX5-25) acceleration and attitude have different cooridinates. Depend on your IMU manufacturer, the two extrinsics for your IMU may or may not be the same. Using our setup as an example:
    1. we need to set the readings of x-z acceleration and gyro negative to transform the IMU data in the lidar frame, which is indicated by "extrinsicRot" in "params.yaml."
    2. The transformation of attitude readings is slightly different. We rotate the attitude measurements by -90 degrees around "lidar-z" axis and get the corresponding roll, pitch, and yaw readings in the lidar frame. This transformation is indicated by "extrinsicRPY" in "params.yaml."
  • IMU debug. It's strongly recommended that the user uncomment the debug lines in "imuHandler()" of "imageProjection.cpp" and test the output of the transformed IMU data. The user can rotate the sensor suite to check whether the readings correspond to the sensor's movement. A YouTube video that shows the corrected IMU data can be found here (link to YouTube).

Citation

Thank you for citing LIO-SAM (IROS-2020) if you use any of this code.

@inproceedings{liosam2020shan,
  title={LIO-SAM: Tightly-coupled Lidar Inertial Odometry via Smoothing and Mapping},
  author={Shan, Tixiao and Englot, Brendan and Meyers, Drew and Wang, Wei and Ratti, Carlo and Rus Daniela},
  booktitle={IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
  pages={5135-5142},
  year={2020},
  organization={IEEE}
}

Part of the code is adapted from LeGO-LOAM.

@inproceedings{legoloam2018shan,
  title={LeGO-LOAM: Lightweight and Ground-Optimized Lidar Odometry and Mapping on Variable Terrain},
  author={Shan, Tixiao and Englot, Brendan},
  booktitle={IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
  pages={4758-4765},
  year={2018},
  organization={IEEE}
}

License

BSD-3-Clause

数据概要
数据格式
IMU, GPS, Point Cloud,
数据量
--
文件大小
15.72GB
发布方
Massachusetts Institute of Technology (MIT)
Founded in 1861 in response to the increasing industrialization of the United States, MIT adopted a European polytechnic university model and stressed laboratory instruction in applied science and engineering. It is frequently regarded as one of the most prestigious universities in the world.
数据集反馈
立即开始构建AI