On Gain Initialization and Optimization of Reduced-Order Adaptive Filter

Despite all the progress in filtering algorithms for state estimation in very high dimensional systems, the technology is delicate and sometimes difficult to apply. Good initialization of filter gain, appropriate choice of tuning parameters and their optimization are the key factors to achieve robust high-performance filtering algorithms. In this paper the authors propose a method for properly initializing the filter gain, and for efficient optimization of the filter performance. Numerical experiments will be given to illustrate the algorithm and to demonstrate the efficiency of the proposed filter for state estimation problems in classical low dimensional as well as in very high dimensional systems.