I know the theory beind ultrasonic and how it is supposed to work. The problem is they are ones of the kind of parts that work mostly in theory, getting them to do the real job is terribly hard. My friends tried to make a displacement/location device using accelerometers and integrating the acceleration in 3D and measuring location as displacement relative to a reference point. The result was less than satisfactory (about 80% error on the average) due to low precision of the accelerometers - and I felt it would be like that from the very beginning when they presented the idea. My feeling about precision of the ultrasonic devices is similar and I would feel nice to be proven wrong on that.

BTW, any experiences with Hall sensors at detecting geographical north?