FILTER FOR POSITION/ATTITUDE ESTIMATION IN UNMANNED AERIAL

State estimation is a very common task in many engineering applications involving dynamic systems. For linear systems corrupted by gaussian noises the problem of optimal state estimation is solved by the well known Kalman filter, which is a finite-dimensional system. In the general case of non-linear/non-gaussian stochastic systems, the optimal state estimator (the Bayesian estimator) is an infinite-dimensional system that provides the conditional probability density of the state given the measurements. Only finite-dimensional approximations of the Bayesian estimator can be implemented in practice, but accurate approximations are computationally prohibitive in most applications. For this reason, many non-optimal filtering techniques have been proposed in the literature, where optimality of the estimate is traded off against lower computational cost. One of the most used estimation algorithms is the so called Extended Kalman Filter (EKF), that is constructed by applying the simple equations of the Kalman filter to the first order Taylor approximation of the equations that define the nonlinear system, computed around the most recent state estimate. The main advantage of the EKF is the simplicity of its construction. However, the use of Jacobians in the EKF has been criticized b