This type of feedback control is called PID control, short for Proportional Integral Derivative. Sometimes it is called a 3 term controller. This is a large subject, but on the other hand there seems to be web sites that include software that implements such control for NXT Segway robots available. See the links below, and maybe follow further links to find something that suits your setup. You may need floating point arithmetic to make the calculation easier. This type of control is also known as a position servo. I am not sure this is the ideal way to do this, but there are examples that do, so that means it works. PID control is widely used in industrial systems.

The PID type of control relies on the error between the set point and the actual position to move the motor that corrects the position. Proportional control closes on the desired position, slowing as it gets closer to the set point, till the motor stops when the error is small. Integral is used to learn from previous history. It is like manually re-adjusting the set point once the robot is positioned, to get closer to the desired position. Derivative anticipates, by determining the position is close to the set point, so the motor can slow and be ready to stop just before it actually gets there. These controls are usually performed on the error signal, the difference between the required position and the actual position. There will be a setting (constant) for each of these 3 terms, which has to be tuned for best operation, such as fastest settling, minimum overshoot and smallest error.

While the second link uses a light sensor to measure distance for vertical, it is not much to translate to angle from a gyro. The set point is of course the angle that is read when the gizmo is balanced, or maybe slightly off balance will work too, with a lot of jiggling around..

If you want to integrate for some other reason, it might be something like adding all the changes in angle (starting from some reference point) to get the current angle, but I would think the gyro has absolute angle outputs which give the current angle. Integration is often used to find "the area under the curve".

The best way to test the sensors is to use the factory programmed sensor tests that came pre-loaded on the NXT block. They tell you which sensor to plug in and then give direct feedback from the sensor (by sound or by displaying something on the screen) so you can see whether its working. If you’ve erased all the programs at any time, though, they won’t be on your NXT.

The link below shows you how to test the sound sensor. On the page are links to directions for testing each of the other sensors.

Future possible improvements:
- Smaller backlash from hardware
- Automatic backlash calibration
- More intelligent driving algorythm
- Continuous sampling, and path adaptation
- Lost line situation handling

Duration : 0:1:6

[youtube C-suay6zTXc]

1. Have 2 thrust fans, and turn one of them off to turn in that direction direction (for ex. turn the left one off to turn left).
2. Have 1 small fan each on the front corners of the hovercraft in addition to 2 fixed thrust fans. Turn them on as needed to shift the hovercraft left and right.
3. Use rudders in the thrust fans to steer the hovercraft.

Which one do u think is the most effective way to steer the hovercraft (and why)? I have a feeling #2 might be best b/c it’s like front wheel drive in cars. #1 and #3 are like rear wheel drive, although #3 appears pretty convincing… what do you think?
In #2, the side fans in the front corner would be facing outward (perpendicular to straight motion), not downward. Just wanted to make this clear.

I would stick with the rudders. Although i have no experience in this matter i do consider myself mechanically inclined. i would be led to believe the front fans would be very clumsy. and the dual fans in the back would be to slow to steer. i look at airplanes,blimps and air boats all using a form of a rudder to steer.

Duration : 0:0:38

[youtube ztkSdgVen-Q]