Revision history [back]

Often you don't need to model the wheel and caster joints to use the URDF with the navigation stack. (They're useful for visualizing the robot, of course.) For example, for a differential-drive robot you need a single frame for the robot base, probably halfway between the two wheels, often called base_link. Then you need frames for each of your sensors, unless you are publishing distances relative to base_link instead. The wheel positions don't really matter – your odometry layer will usually publish the transform between the odom frame and the base_link frame.

The transforms from base_link to the sensor frames are usually published automatically by robot_state_publisher and joint_state_publisher. Movable joint transformations are handled the same way, but you have to publish the joint positions so that joint_state_publisher can determine the right transformations.