New approaches to improve estimator performance for quadrotors

In order to use Kalman filter effectively, covariance matrices of process and measurement noise must be known a priori. Nevertheless, these values may not be known exactly which causes the filter to work under suboptimal and even in divergent conditions. In most cases, these parameters are guessed or tuned by trial and error approach both of which do not guarantee optimality and convergence. To ensure near - optimal conditions for the measurement noise, a bio-inspired artificial bee colony optimization algorithm based Kalman filter offline tuning scheme is introduced. In addition, an objective function is proposed to handle the convergence problem of the existing function. Mathematical proofs are derived for the convergence problem and also for proposed function to define its behavior on the search space. Simulation outputs are compared with genetic algorithm based Kalman filter for both objective functions. Since optimization of both process and measurement noise covariance matrices is a challenging task in literature, an evolutionary algorithm based Kalman filter is proposed to simultaneously estimate the process and measurement noise covariance matrices of the Kalman filter to improve the performance of the sub – optimal filter. A surrogate – assisted fitness function is also introduced to achieve multi – dimensional simultaneous optimization with finite time consideration. Results are compared with an optimal Kalman filter by means of absolute error, root mean square error, and mean absolute error. Efficacy of the proposed algorithms and functions are shown according to the performed simulations and numerical results of the Monte Carlo simulations.