Adaptive Mid360？

[Wangshihu12]

Hello, excellent project.
Is this algorithm suitable for Mid360, or how to adjust the parameters for Mid360?

[LC-Robotics]

Yes, this algorithm is suitable for almost all LiDARs. Since the algorithm takes a pointcloud and a pose as input, you also need a SLAM algorithm to provide the pose for the pointcloud, such as Fast-LIO. Fast-LIO outputs /cloud_registered_body (the undistorted point cloud in the IMU frame) and /tf during runtime. You only need to modify the following parameters in FreeDOM's launch file: pointcloud_topic, map_tf_frame, and sensor_tf_frame to match the Fast-LIO outputs:

<launch>
    <node name="freedom" pkg="freedom" type="freedom_node" output="screen">
        <rosparam ns="/freedom" file="$(find freedom)/config/indoor_stairs.yaml" command="load" />
        
        <param name="pointcloud_topic" type="string" value="/cloud_registered_body" />
        <param name="map_tf_frame" type="string" value="camera_init" />
        <param name="sensor_tf_frame" type="string" value="body" />
        <param name="save_map_topic" type="string" value="/save_map" />
        <param name="save_map_path" type="string" value="$(find freedom)/generated_pcd/" />
        <param name="fov_mask_path" type="string" value="$(find freedom)/config/" />
        <param name="learn_fov" type="bool" value="false" />
        <param name="enable_visualization" type="bool" value="true" />
    </node>

    <node type="rviz" name="rviz" pkg="rviz" args="-d $(find freedom)/config/rviz_indoor.rviz"/>
</launch>

Note that in Fast-LIO, you can set point_filter_num to 1 to prevent downsampling and achieve better mapping results.

The above steps are effective for other common LiDARs as well. Since the IMU and LiDAR center in the MID360 are very close and their coordinate frames are aligned, you can approximate the body frame as the LiDAR position. If the IMU and LiDAR are farther apart or their coordinate frames are not aligned, you can add a static TF in the launch file to connect the IMU frame and the LiDAR frame, and set sensor_tf_frame to the LiDAR frame.

The indoor_stairs.yaml file contains the parameters I used for MID360 in the Indoor dataset, which should work. If you have any further questions, feel free to ask me.

[Wangshihu12]

Thank you for your detailed answer. Sorry for replying so late. I will try the effect of Mid360 as soon as possible. It's a pleasure to communicate with you.