Home > Publications > Doctoral dissertation & Master thesis
Domestic Conference
Title Dynamic State Estimation for Multi-platform and Multi-sensor Fusion
Degree MS
Author Jaeuk Byun
Advisor Kiseon Kim
Graduation Date 2014.08.25 File
    Date 2017-03-06 10:35

Wireless Sensor Network (WSN) has gathered an enormous attention with monitoring and tracking in military, industry, and civil application. State estimation technology to approximate the system state from sensor observation is a main interest of WSN, and is essential in data fusion. As the fusion in estimate fusion is achieved based on information at local platforms and sensors, the data processing capability from a series of sensor measurements is important. The parametric/nonparametric methods have been introduced for dynamic sensor estimation and fusion. Among dynamic state-space model based algorithms, we conclude that the parametric approach, especially Unscented Kalman filter, is suitable for dynamic state estimation, whose on-line processing of nonlinear information is essential as in battlefield surveillance and missile tracking.

Unscented Transform (UT) used in UKF approximates nonlinear transformation for Kalman filtering with a few number of deterministic samples named sigma point, but the estimation performance with UT can be very erroneous due to no consideration of nonlinear transform in the selection of sigma point. This thesis suggests a point update method to control the weight of specific sigma points by moving the location of UT sigma points. The numerical result shows that with a proper selection on nonlinear transform and parameters, the suggested idea can reduce the error of UT. Further consideration of update rule and parameter selection for the concrete and general results remained as the further work of this research.

In addition, the simulation of tracking tactical target with selected motion model is performed to examine the effect of DSS-based filtering in a military application. From the simulation, we conclude that a filtering method covering the abrupt state changes is required for tracking in military application.

광주과학기술원 한·러 MT-IT 융합기술연구센터 광주과학기술원정보통신공학부