Revision history [back]

If you look at Coordinate Frames and Transforming Sensor Data you'll notice that pose data is interpreted relative to the world frame, while twist data is interpreted relative to the base_link frame.

One proper way of providing the data would thus be to compute the velocity from the encoder ticks and time between encoder reads and then provide a nav_msgs/Odometry message with the correct velocity set in the twist/twist/linear/x field. This can then be fed into robot_localization (the correct field has to be set to "true" for the config of course).