Integração GPS/INS baseada em filtro de Kalman para navegação

Abstract

Navigation means determining position and attitude of a moving body in relation to a reference over time. The quantities position, velocity and attitude, which allow this body to be georeferenced, are called navigation states. The combination of sensors that makes it possible to determine navigation states is known as a navigation system. The integration of navigation systems seeks to overcome the quality of states estimation in relation to that obtained with any of the systems operating in isolation. For this, systems with complementary characteristics are typically combined. In this context, the present paper involved the development of a navigation system using GPS/INS integration through two Kalman filters. The first is used for attitude estimation, using accelerometer, gyroscope and magnetometer readings. The second is used to estimate the position from the GPS measurement and the accelerometer output. GPS/INS integration benefits from small short-term INS estimation errors and the fact that GPS measurements do not diverge over time. To validate the implemented system, a reference trajectory was defined, from four points that were defined using a precision GPS. Such validation was performed in simulation and from experimental data. With the simulated data, the GPS/INS integration presented a better result than the GPS measurement alone. More precisely, the mean error and the error variance of the values estimated by the filters were smaller than those of the measured values. However, in the test with real data, the results of the developed navigation system were below expectations. This was due to the high noise level of the low-cost accelerometer used, in line with the double integration performed in the dynamic model of the Kalman filter for position estimation.