Abstract:
The invention of Kalman filter brings to the engineer world an important solution on the noise management, and more in general, the possibility of smoothing perturbed data collected from a sensor.
So far, studies and upgrades adopted in this environment helped researchers to develop different kind of filters, adopting different methods and mathematical tools which allowed to handle different kinds of data and perform better.
The aim of this thesis is study the most used Kalman filters, and in particular the combination of them with algebraic structures such as Dual Quaternions, in order to establish, given a set of data coming from an UAV model, which of them performs better, giving the best approximation on results gained from measurements and getting the best ideal trajectory.
Data are withdrawn from an FPV drone, by recording a flight session and saving logs in the built-in black-box. These are then parsed, transformed in a csv file and only relevant data, such as axis rotations, GPS coordinates and altitude, are used to perform our experiments.
Several plots were generated to help us to made a comparison about results obtained from our study, concluding that the Unscented Kalman filter with Dual Quaternions is the best choice to dealing with these kind of data.
