> Could<p>Yeah. Precision of accelerometers is a tough problem. Dealt with some on my engineering degree and because they measure the second derivative of position we can add 'integrate' it twice to get the position, but that process yielded terrible results after centimeters. The imperfect nature of accelerometers adds up a lot of errors along the way. An entirely theoretical, infinitely sensitive, and perfect accelerometer would give us the perfect position if we know what is our starting position and although I don't have any experience with a "quantum" accelerometer, I don't think a quantum accelerometer is perfect enough and I don't think this article made a good case about it.<p>I thought of a silly project though: What if we tried to fix the problem of customer-grade accelerometers being too imprecise by parallelizing the accelerometers?<p>We would connect a large number of sensitive accelerometers, something like 100 or 1000, maybe of different frequencies, and feed a segment of the sensory data over time to a neural network measuring the fitness of it towards minimizing the prediction error for long distance traveling such as needed by ships or anything really. Does anyone know if any project has been attempted in the past?