Discrete adaptive filters for short-lived dynamic systems

Abstract

Reliable and easily implemented discrete adaptive filters for short-lived systems are identified. Three different filtering techniques and a lightly damped dynamic system are used to illustrate boundaries specified by the convergence criterion. The filtering techniques in this study are the modified extended Kalman filter, the decoupled Kalman filter, and a pseudolinear regression filter. The extended Kalman filter is shown to converge once it identifies the appropriate decoupling of states and parameters. The decoupled Kalman filter provides a much cleaner convergence but has the standard computational burden of the Kalman filter as well as possible convergence problems. The pseudolinear regression algorithm provides excellent convergence with a much more computationally compatible and time-sensitive algorithm.