Re: misalignment from tango multi session

Posted by matlabbe on
URL: http://official-rtab-map-forum.206.s1.nabble.com/misalignment-from-tango-multi-session-tp5046p5752.html

Hi,

Voxel size of 2 meters seem high to me as it is 20 cm in the velodyne example. If you want to save time and memory, uncomment those lines:

<!-- param name="Mem/LaserScanVoxelSize"     type="string" value="0.1"/ -->
<!-- param name="Mem/LaserScanNormalK"       type="string" value="10"/ -->
<!-- param name="Mem/LaserScanRadius"        type="string" value="0"/ -->

and comment this one under rtabmap node:

<param name="Icp/VoxelSize"                  type="string" value="2"/>

In that setup you don't seem using the IMU (use_imu argument is false), I recommend to use one if available. For the transform IMU->lidar, the orientation should be accurate. For example, make sure the imu frame matches the lidar frame.

The odometry seems drifting quite a lot, can you share a rosbag (lidar, tf, imu)? It would be easier to debug.

You can use a kinect at the same time. The rgbd_sync approach is indeed to way to go. The warning here is that rgbd_sync didn't receive any kinect data. Verify kinect topics are published:

$ rosrun topic hz /camera/rgb/image_rect_color /camera/depth_registered/image_raw /camera/rgb/camera_info

cheers,
Mathieu