Sensor fusion with cointegration analysis for IMU in a simulated fixed-wing UAV
Elias de Souza Goncalves, Paulo Fernando Ferreira Rosa · 2017
This work deals with the problem of navigation and location of an unmanned aerial vehicle (UAV) that results in the estimation of the state variables of the vehicle. In order to solve the problem, first we chose a fixed-wing aircraft model available for flight simulators. Then, we applied the cointegration method to a set of inertial sensors composed of two accelerometers, and two gyroscopes. It is an analytical technique to verify common trends in multivariate series and long-term and short-term dynamic modeling. That allows us to discover the best readings from one IMU, or if not possible, we can find out the best features of each IMU working together. Thus, the inherent contribution of this work is the use of cointegration as a way of estimating the behavior of the UAV inertial sensors. In the last step, a widely used tool for the prediction of state variables, the Extended Kalman Filter (EKF), merges the sensors of the previous step with the GPS to eliminate inaccuracies. A software in the loop architecture is proposed as a validation methodology. The result shows that the estimates of the state variables were satisfactory, always remaining close to those considered true and calculated by the embedded software.