Discrete-Time Linear Filtering in Arbitrary Noise W.pdf

Discrete-Time Linear Filtering in Arbitrary Noise
X. Rong Li University of New Orleans, xli@uno.edu Chongzhao Han and Jie Wang Xi’an Jiaotong University, czhan@xjtu.edu.cn Abstract—The Kalman filter is a recursive Best Linear Un- biased Estimator (BLUE) for a linear dynamic system with uncorrelated white process and measurement noises. It has been extended to the case where the noises are Markov and/or crosscorrelated for the same time instant. This paper presents optimal batch and semi-recursive filters and a suboptimal re- cursive filter for a linear discrete-time system with arbitrarily colored (not necessarily Markov) noises that are arbitrarily cross-correlated and correlated with the initial state of the system. They are generalizations of the Kalman filter for the case of arbitrary additive noise of known first two mo- ments. Numerical examples are provided. They demonstrate the superiority in terms of performance and efficiency of the proposed recursive filter. 1 Introduction The development of the Kalman filter [4] is widely regarded as one of the breakthroughs that marked the beginning of the era of modern control system theory [8]. It has found wide- spread applications in many areas. The Kalman filter pro- vides an optimal recursive solution to the problem of filtering the state of a linear dynamic system, which has the following two major advantages over the Wiener filter: it is in a very simple recursive form in time domain and is perfectly suitable for nonstationary as well as stationary processes. Although optimal only for linear systems, the Kalman filter plays an important role for many nonlinear problems. For example, it is the backbone of most solutions for nonlinear state es- timation where the nonlinearity is mild; it often serves as a building block for many more sophisticated state estimators, such as the multiple-model method, a state-of-the-art approach for many hybrid estimation problems [1, 5]. The original Kalman filter is valid only f