Dear ros users,

 

probably we figured out a bug in the robot_state_publisher (or are not using it correctly):

 

We are running a robot_state_publisher subscribed to the /joint_states topic and using a urdf robot_description. The transformation from joint_states with urdf to tf transformations works only if

a)       the joint information for the whole kinematic chain is published in one joint_states message or

b)       at least there are no joints connected to a serial chain and published in different joint_states messages.

 

In our case e.g. there is a gripper attached to an arm. Both have rotary joints defined and are publishing joint_states in two different joint_states messages, but tf can only be computed for the arm, not for the gripper.

 

Is this a known behaviour (bug?) or are we missing something?

 

Best regards,

Florian