Estudo de aplicação de filtragem estocástica ao problema de rastreamento de trajetórias de um drone

Abstract

Stochastic filters can be used in several situations, as in unmanned aerial vehicles tracking problems. In this case, filtering is used for estimating trajectories and other relevant parameters, such as velocity, from measurements provided by position sensors, which are noisy. This work studies the Kalman, Extended Kalman and IMM filters applied to different trajectories, created through a simulator based on the Frenet-Serret equations. The trajectory simulator implemented proved to be a very useful tool in the stochastic filters study and application, being possible to create several trajectories using few parameters. The results obtained through the filtering performed by each filter were compared to each other using the the mean squared error, which made possible to observe: that the KF has a better performance than the EKF when dealing with a curve that is described as linear system; the opposite is truth when dealing with a curve described by a non-linear system; the IMM has better performance than both, even in their field of expertise due to the switching that it’s able to do between the filters that can better deal with the curve in hand.