Sorry, Please let me tell a small story.

I am new to ROS, my final goal is "my robot can build a map automaticlly and walk to any demanded position". maybe in professional terms, what I want is SLAM and Navigation. a basic goal being achieved lots of times by most ROSers, isn't that true?

• I tried stereo cameras + rgbdslam at first. but rbgdslam often stops, cannot build a 3D map smoothly.
• I realized that maybe 2D SLAM is enough for me. so I tried kinect + gmapping. but I don't have a wheel encoder or other odometery device. the map generated is limited within a small fixed angle.
• Someone told me to use hector_mapping in place of gmapping, which need no odometery input. I did it. but the result map jumps while kinect is turning around. I think the reason might be the laser scan data simulated by kinect is short sighted with noises.

I think now I have 4 ways to try:

1. Use a real laser scan sensor, with hector_mapping. but laser sensor seems to be very expensive($500?)
2. Can I use a ultrasonic sensor instead of a real laser scan sensor? someone said yes here(answer.ros), but I didn't get it.
3. Use a wheel encoder, with gmapping. but how to drive the encoder and publish odom frame to /tf ? is there propery driver I can use(I searched but did not find), or must I write my own one?
4. Is there anybody who can help me achieving my goal online via some tool like skype? I am losing confidence and patient on my own. I will pay for that.

If you are familiar with 3(wheel encoder solution), could you show me some tutorials about how to program a driver to satisfy gmapping's /tf topic? sample is best(I have read http://wiki.ros.org/navigation/Tutorials/RobotSetup/Odom).

Or, maybe I am walking on the wrong way from my objective. any hints are welcome. many many thanks. ROS is too difficult for me :(