Analytical techniques for tracking filter implementation

Daniel Gleason, Purdue University

Abstract

The objective of this study is the development and verification of nonlinear and linear analysis techniques for aircraft tracking filter implementation. The aircraft tracking problem has two distinct requirements. The first requirement is to accurately estimate the position of the aircraft at a present point in time. The second requirement is to predict the position of the aircraft at a given future time. Presently implemented tracking filters suffer from estimation and prediction accuracy degradation from two causes. The first cause is the lack of measurement information concerning the aircraft acceleration state. The second cause of error is attributable to a modeling insufficiency since, typically, the tracking filter is provided with no information concerning the system inputs. This study addresses these two fundamental shortcomings to currently employed tracking filters. First, a nonlinear analysis is presented. In this analysis, a tracking filter that incorporates aircraft orientation in both the system model and measurements is compared to a tracker that uses only standard radar measurements. The addition of the orientation information enhances the tracking filter performance because, in general, aircraft orientation is strongly correlated with the acceleration of the aircraft. Improved position estimation and prediction is demonstrated with the tracking filter that incorporates orientation information. The second fundamental cause of tracking filter errors (i.e. unknown system inputs) is investigated using linear analysis techniques. First, an analysis technique is developed using frequency and time domain methodologies for determining tracking filter gains for a system with unknown exogenous inputs. An analysis technique is also developed to determine the difference in tracking filter error covariance histories using variable structure models. This leads to a means for calculating process noise levels to achieve error covariance equivalent models. In addition, a three state, constant gain discrete tracking filter is investigated to determine filter gain behavior as a function of process and measurement noise levels, and measurement sampling times. A limiting analysis is presented to determine the asymptotic behavior of the filter gains.

Degree

Ph.D.

Advisors

Andrisani, Purdue University.

Subject Area

Aerospace materials

COinS