Just a warning ahead: this is going to be a lot of work.
Step 1: controlling the arm from code
You need a way to control the motors from code, ideally under Linux (since ROS is best-supported under Linux). Luckily for you, somebody already has reverse-engineered the USB protocol, so you will probably be able to use that.
First, I would write a small test program that can move any motor of the arm to a desired joint angle and publishes JointState messages for each joint on the /joint_states
topic (ignore the gripper for now, that's going to be easy later). I'm not sure how well that's going to work, since the arm doesn't measure the joint angles, and the only available commands seem to be "start/stop movement in some direction". You'll probably have to time the commands and compute the resulting joint angles.
Step 2: create an URDF model of the arm
This can be done in parallel to the first step; follow the URDF tutorials. It would be great if you had CAD models of your arm, but for the beginning just using some cylinders/boxes instead should do. Ideally, after completing both steps, you should be able to control the joints of the arm and visualize the result in RViz. You could even point a Kinect camera at your arm, add that point cloud to RViz and check how well the actual arm movements and the URDF model match up.
Step 3: arm navigation
The first two steps should have kept you busy for some while. :-)
By now, just by using the joint angles, you should already be able to do some interesting stuff with the arm. Also, you should have a feeling for the arm, and what can be done with it and what not. If you want to go further, you could try to follow this tutorial to get arm navigation running. This would enable you to pick and place objects recognized in a point cloud (e.g. from a Kinect) while avoiding obstacles. However, this is going to be a real lot of work, and I'm not sure if your arm is maybe too inaccurate (especially when the joint angles can only be computed using some dead reckoning) to do that. Perhaps you need to go for some simpler solution.
Feel free to ask more questions. I'm really looking forward to see how your project is going.