Intelligent adaptive unscented particle filter with application in target tracking
- PDF / 2,340,640 Bytes
- 9 Pages / 595.276 x 790.866 pts Page_size
- 36 Downloads / 254 Views
ORIGINAL PAPER
Intelligent adaptive unscented particle filter with application in target tracking Ramazan Havangi1 Received: 15 June 2019 / Revised: 4 February 2020 / Accepted: 18 March 2020 © Springer-Verlag London Ltd., part of Springer Nature 2020
Abstract The particle filter (PF) perform the nonlinear estimation and have received much attention from many engineering fields over the past decade. However, the standard PF is inconsistent over time due to the loss of particle diversity caused mainly by the particle depletion in resampling step and incorrect a priori knowledge of process and measurement noise. To overcome these problems, intelligent adaptive unscented particle filter (IAUPF) is proposed in this paper. The IAUPF uses an adaptive unscented Kalman filter filter to generate the proposal distribution, in which the covariance of the measurement and process of the state are online adjusted by predicted residual as adaptive factor based on a covariance matching technique. In addition, it uses the genetic operators to increase diversity of particles. Three experiment examples show that IAUPF mitigates particle impoverishment and provides more accurate state estimation results compared with the general PF. The effectiveness of IAUPF is demonstrated through Monte Carlo simulations. The simulation results demonstrate the effectiveness of the proposed method. Keywords Particle filter · Target tracking · Adaptive unscented Kalman filter · Genetic operators
1 Introduction The problem of nonlinear state estimation is to estimate the possible dynamic state of a nonlinear system, based on a set of noisy observations [1, 2]. It plays an important role in many fields, such as, statistical signal pro cessing, communication, tracking [3], navigation [4], and so on. The PF is an effective estimator for the nonlinear/non-Gaussian systems [5]. It is a sequential Monte Carlo Bayesian estimator that constructs probability density function (PDF) using a set of random particles with associated weights [5]. In order to improve the performance of PF, choosing of the proposal distribution and the resampling step is importance [6]. In classical PF, the state transition is often chosen as the proposal distribution [1, 7]. PFs using transition prior as the proposal distribution is named as bootstrap filter [8]. Although bootstrap filter has achieved great success in many applications, its main drawback is that it does not include information of the new observations when sampling from
B 1
Ramazan Havangi [email protected] Faculty of Electrical and Computer Engineering, University of Birjand, Birjand, Iran
the proposal distribution [8, 9]. In addition, a property of particle filter is that the variance of sample weights increases over time [10]. To solve this problem, the resampling step is performed [11]. In the resampling step, the particles with large weights are replicated and the ones with negligible weights are removed [12, 13]. This lead to a great loss of diversity of particles and thus bring another problem that call
Data Loading...
