Hi everyone,

So as a part of my GSoC 2018 project, I am using the Kinect Fusion algorithm for the process of 3-D reconstruction. Initially, I am launching the robot in Gazebo and the robot in RViz! with MoveIt!. I have placed a random sphere object for visualization in Gazebo.

After this, I launch the Kinect Fusion reconstruction node which utilizes a TSDF based approach for Fusion using the YAK package. However, as soon as this file is launched, after a time of 2-3 seconds, I get this one warning on the RViz console.

[ WARN] [1531932207.525024167, 3837.520000000]: No transform available between frame 'map' and planning frame '/world' (Could not find a connection between 'world' and 'map' because they are not part of the same tree.Tf has two or more unconnected trees.)

The Kinect fusion node console shows the following:

[ INFO] [1531932206.867190280]: starting kinfu node...
Device 0:  "GeForce GTX 960M"  4044Mb, sm_50, 640 cores, Driver/Runtime ver.9.0/8.0
[ INFO] [1531932206.886842801]: Volume size is 1.568 0.768 0.384
[ INFO] [1531932207.009896212]: Fixed frame: /map Camera frame: /camera_depth_frame
[ INFO] [1531932207.029919344]: Use pose hints set to 1
[ERROR] [1531932207.172164447, 3837.196000000]: Lookup would require extrapolation into the past.  Requested time 3837.187000000 but the earliest data is at time 3837.191000000, when looking up transform from frame [camera_depth_frame] to frame [volume_pose]
[ INFO] [1531932207.172317940, 3837.196000000]: Connecting to camera...

[ INFO] [1531932207.226409163, 3837.246000000]: Connected.

bilateral_sigma_depth: 0.04
bilateral_sigma_spatial: 4.5
gradient_delta_factor: 0.5
icp_angle_thresh: 0.523599
icp_dist_thresh: 0.1
icp_truncate_depth_dist: 0
raycast_step_factor: 0.05
tsdf_min_camera_movement: 0.0001
tsdf_trunc_dist: 0.005
volume_resolution: 0.001
[ INFO] [1531932207.247908859, 3837.264000000]: volPos (loaded): 0, 0, 0
[ INFO] [1531932207.320968639, 3837.324000000]: Volume size set to: [1.568, 0.768, 0.384]
[ INFO] [1531932207.321760993, 3837.324000000]: Volume pose set to: [1, 0, 0, 0;
 0, 1, 0, 0;
 0, 0, 1, 0;
 0, 0, 0, 1]
Reset Volume
[ INFO] [1531932207.330725281, 3837.332000000]: Starting tracking...

[ERROR] [1531932207.969904398, 3837.948000000]: Could not find a connection between 'volume_pose' and 'camera_depth_frame' because they are not part of the same tree.Tf has two or more unconnected trees.
[ INFO] [1531932207.972523366, 3837.948000000]: 
Motion Input: [0, 0, 0, 0;
 0, 0, 0, 0;
 0, 0, 0, 0;
 0, 0, 0, 1]
Motion Corrected: [0, 0, 0, 0;
 0, 0, 0, 0;
 0, 0, 0, 0;
 0, 0, 0, 1]
Pose Input: [0, 0, 0, 0;
 0, 0, 0, 0;
 0, 0, 0, 0;
 0, 0, 0, 1]
Pose Corrected: [0, 0, 0, 0;
 0, 0, 0, 0;
 0, 0, 0, 0;
 0, 0, 0, 1]

The same Octomap is flipped to the right side. This happens consistently on n number of trials.

I have read in some posts that a static publisher should be used in order to avoid the error.

The Kinect fusion launch file is creating a static transform publisher like this.

<node name="map_to_volume_publisher" pkg="tf" type="static_transform_publisher" args="0 0 0 0 0 0 volume_pose map 1000" output="screen"/>

While the robot TF is published in the moveit_planning_execution.launch file as follows when MoveIt! amd Gazebo was launched on the start.

  <!-- publish the robot state (tf transforms) -->
  <node name="joint_state_publisher" pkg="joint_state_publisher" type="joint_state_publisher"/>

This is how the world and map frames look like as visualized in RViz. The map frame is with the volume_pose frame near the base of the robot, while the world frame is on the floor.

I would be grateful for any suggestions on how to remove this discrepancy as I feel this could potentially affect the Fusion process ahead. Is the octomap created due to loss in tracking by the camera when the Fusion process has started?

Thanks in advance.

System info: I have used Ubuntu 16.04 and ROS Kinetic