I attached a robotic arm to my turtlebot. I managed to run moveit + pick and place demo (which is picking up an imaginary object with the arm).

I tried the next step which is picking up real objects. The block manipulation demo is kind of running (randomly crashing or doing nothing 70% of time but sometimes I can click and place a block) but obviously the arm is moving to a completely different position. So I tried running the kinect_calibration. Currently I am only using one kinect (which was used for navigation), moving it around to get a good view. (I plan on buying a second kinect and adding it off-board)

Basically I tried exactly the same as in this video link. The generated output (tf static transform (which should be used for calibrating the kinect off-board)) is doing nothing at all when executed. The XML commands for the on-board setup did not work, too. Then I managed to find a file (turtlebotproperties.urdf.xacro) which contains similar XML statements (`<property name="campx" value="0.183127">`). By changing these it kind of works.

So I wonder how all of this will work with a second kinect (off-board). The camera picture in RVIZ always spawns in front of the kinect, so would there be the need of adding a second kinect to the URDF model aswell? But is this even possible for an off-board kinect?

Secondly I wonder how it is possible to add a second kinect to ROS. Wouldn't they publish on the same topics? And how to tell the block manipulation which one to use?

Unfortunately none of this is described in the turtlebot_arm (calibration) tutorial (which seems to be very outdated anyway)

I am using ROS indigo, corb555 arbotix fork and corot turtlebot_arm fork (which offer support for the phantomX pincher arm).