This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS: SYSTEMS 1
Linear and Nonlinear Regression-Based Maximum
Correntropy Extended Kalman Filtering
Xi Liu , Zhigang Ren , Member, IEEE, Hongqiang Lyu , Zhihong Jiang, Pengju Ren, Member, IEEE,
and Badong Chen
, Senior Member, IEEE
Abstract—The extended Kalman filter (EKF) is a method
extensively applied in many areas, particularly, in nonlinear tar-
get tracking. The optimization criterion commonly used in EKF
is the celebrated minimum mean square error (MMSE) crite-
rion, which exhibits excellent performance under Gaussian noise
assumption. However, its performance may degrade dramatically
when the noises are heavy tailed. To cope with this problem,
this paper proposes two new nonlinear filters, namely the linear
regression maximum correntropy EKF (LRMCEKF) and non-
linear regression maximum correntropy EKF (NRMCEKF), by
applying the maximum correntropy criterion (MCC) rather than
the MMSE criterion to EKF. In both filters, a regression model
is formulated, and a fixed-point iterative algorithm is utilized to
obtain the posterior estimates. The effectiveness and robustness
of the proposed algorithms in target tracking are confirmed by
an illustrative example.
Index Terms—Extended Kalman filter (EKF), fixed-point algo-
rithm, maximum correntropy criterion (MCC).
I. INTRODUCTION
T
ARGET tracking and positioning is sufficiently necessary
in many fields, including orbit determination and
autonomous navigation in satellite network, mobile location
estimation in communication network, moving target track-
ing in wireless sensor network, and many others [1]–[6]. It
is often very important to estimate the position, velocity, and
some other useful quantities through noisy measurements from
a moving target. For linear dynamic systems, the Kalman fil-
ter (KF) is an effective tool for solving this problem, which
is a recursive least squares linear filter [7]–[10]. For nonlinear
Manuscript received January 20, 2019; revised February 28, 2019; accepted
May 4, 2019. This work was supported in part by the 973 Program under
Grant 2015CB351703, in part by the National Natural Science Foundation
of China under Grant 61873199 and Grant 91648208, and in part by the
National Natural Science Foundation-Shenzhen Joint Research Program under
Grant U1613219. This paper was recommended by Associate Editor M. Basin.
(Corresponding authors: Zhihong Jiang; Zhigang Ren.)
X. Liu, H. Lyu, P. Ren, and B. Chen are with the School of Electronic
and Information Engineering, Xi’an Jiaotong University, Xi’an 710049,
China (e-mail: lx1102@stu.xjtu.edu.cn; hongqianglv@mail.xjtu.edu.cn;
pengjuren@mail.xjtu.edu.cn; chenbd@mail.xjtu.edu.cn).
Z. Ren is with the School of Electronic and Information Engineering,
Xi’an Jiaotong University, Xi’an 710049, China, and also with the School of
Mathematics and Statistics, Xi’an Jiaotong University, Xi’an 710049, China
(e-mail: renzg@mail.xjtu.edu.cn).
Z. Jiang is with the School of Mechatronic Engineering, Beijing Institute
of Technology, Beijing 100081, China (e-mail: jiangzhihong@bit.edu.cn).
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TSMC.2019.2917712
systems, we usually apply some nonlinear extensions of the KF
to estimate the states. Well-known examples include extended
KF (EKF) [11], unscented KF (UKF) [12], [13] and cubature
KF (CKF) [14]. EKF is a widely used nonlinear Kalman type
filter, which approximates the nonlinear equation by its first-
order linearization and then uses the original KF procedure to
complete the filtering process. UKF is derived based on the
unscented transformation and CKF is based on the third-degree
spherical-radial cubature rule. However, the optimization cri-
terion used in these filters is the celebrated minimum mean
square error (MMSE) criterion, which is sensitive to heavy-
tailed noises and may lead to poor performance [15]. The
heavy-tailed noises occur in many real-world applications, for
example, radar measurement system and GPS measurement
system.
To cope with the non-Gaussian filtering problems, some
robust methods are proposed. The Huber methodology
uses the Huber-based cost function to improve the robust-
ness [16], [17]. The student’s t filter supposes the process and
measurement noises obey student’s t distribution [18], [19].
Besides them, over the past few years, the optimization
criteria in information theoretic learning (ITL) [20] have
received much attention from many researchers, which
adopt the information theoretic quantities estimated directly
from the data. As one of the information theoretic quan-
tities, the correntropy can capture higher-order statistics
of errors and has recently been utilized as a robust cost
to provide significant performance improvement in many
fields [21]–[34]. In particular, the optimization criterion with
respect to correntropy is called the maximum correntropy
criterion (MCC) in ITL [20]. It has been widely used
in robust adaptive filters against heavy-tailed non-Gaussian
noises [20], [22]–[25], [29].
We often use some iterative algorithms to obtain the MCC
solution. The commonly used ways are the gradient-based
methods [22]–[25], which are simple and easy to understand.
However, they involve a free parameter, i.e., step-size, and
their convergence speeds are slow. An alternative effective way
is the fixed-point iterative algorithm, which contains no step-
size and may have a fast convergence speed [20], [35], [36].
Specifically, a sufficient condition on convergence charac-
teristics of the fixed-point MCC algorithm was presented
in [37].
In recent works, some linear KFs based on MCC [38]–[41]
were proposed, and some correntropy-based UKFs [42]–[45]
were developed to solve the problems of nonlinear systems.
2168-2216
c
2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.