Revision history [back]

The usual way in path planning and navigation is to create velocity commands for the full robot. You usually apply the velocities to the base_link frame.

Then, you need a base controller which computes the given wheel velocities given the "full robot velocity" depending on you kinematics. Thus, the /cmd_vel topic usually goes into the base controller, which then computes the wheel velocities.

Once you have the kinematic model of the robot, which you require to compute the command velocities for each wheel, you can use the inverse of this to compute the odometry from the current wheel velocities and then publish the odometry to the ROS system.