2-Lidar and 3-Lidar hybrid

[jacklu333333]

Hi,
I am using an intel realsense L515 lidar camera with rtabmap on my turtlebot2.
However, I am experiencing serious inaccuracy of the odometry at the mapping sessions due to the rotation error accumulation.
I was wondering that whether is it possible to use additional lidar to provide the odometry so that it can maintain the x-y plane orientation even after several rotations?
(Note that I am using the pointcloud for 3-D mapping)

Thanks advance.
Jack Lu

[mtal]

Hello Jack Lu, you can use this tutorial to set it up on your Turtlebot2: SetupOnYourRobot Take a look on a 2.1 since your L515 sensor is producing depth data similar like Kinect camera. Wheel odometry can be used as well. Additionally you can select which data you want to use to build the 2D map by setting this parameter:

<param name="Grid/FromDepth" type="string" value="false"/>

If set to false, a 2D Lidar is going to be used to build the 2D map. Regards

[jacklu333333]

Hi, thank you for your reply.
In the link, you share it is still 2-D mapping with additional 3-D information. The shape of the object will be defined by the scan. In this case, irregular shape objects like chairs will not be well-identified.

Hence, what I want is to have a 3D mapping with suscribe_cloud. However, my camera only has 77 degrees for the x-y plane.
Therefore I want to use the 2-D lidar scan to provide odometry for the accuracy of the x-y plane. (It perform poorly with only the lidar camera)

Is there any way I can achieve this?

[matlabbe]

Hi,

You can use subscribe_scan for your 2d lidar, then enable RGBD/NeighborLinkRefining while subscribe to wheel odometry, similar to this setup: http://wiki.ros.org/rtabmap_ros/Tutorials/SetupOnYourRobot#Kinect_.2B-_Odometry_.2B-_2D_laser, the wheel odometry will be corrected by the lidar.

To generate a map from the L515, set

<param name="Grid/FromDepth" type="string" value="true"/>

cheers,
Mathieu