A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Unscented Kalman filter for long-distance vessel tracking in geodetic coordinates
https://orcid.org/0000-0002-4957-6903
https://orcid.org/0000-0002-4188-9614
Abstract This paper describes a unique instantiation of the unscented Kalman filter (UKF) that is particularly well-suited to long-distance surface vessel tracking. The proposed filter leverages a nonlinear process model to predict the position, speed, and heading of targets directly in geodetic coordinates, obviating the need for intermediate coordinate frame transformations, and enabling the use of a simple linear observer. These features differentiate the proposed filter from prior architectures, most of which require the definition of a planar coordinate system in order to describe the linearized state kinematics in an easily differentiable form. Reliance upon local Earth-tangent frames is acceptable in confined operating regions, but incurs substantial estimation error when the target is far from the local origin. It is shown that this error grows with distance-cubed from the local origin. By tracking targets directly in geodetic coordinates, the proposed filter is able to achieve more consistent performance, and a three-fold reduction in computational cost. A series of numerical simulations and field exercises are performed to ground these claims, and verify the efficacy of the proposed filter.
Highlights The proposed filter enables target tracking directly in geodetic coordinates. State update equations are derived from 1st principles, using spherical trigonometry. Performance is shown to match or exceed that of a comparable extended Kalman filter.
Unscented Kalman filter for long-distance vessel tracking in geodetic coordinates
https://orcid.org/0000-0002-4957-6903
https://orcid.org/0000-0002-4188-9614
Abstract This paper describes a unique instantiation of the unscented Kalman filter (UKF) that is particularly well-suited to long-distance surface vessel tracking. The proposed filter leverages a nonlinear process model to predict the position, speed, and heading of targets directly in geodetic coordinates, obviating the need for intermediate coordinate frame transformations, and enabling the use of a simple linear observer. These features differentiate the proposed filter from prior architectures, most of which require the definition of a planar coordinate system in order to describe the linearized state kinematics in an easily differentiable form. Reliance upon local Earth-tangent frames is acceptable in confined operating regions, but incurs substantial estimation error when the target is far from the local origin. It is shown that this error grows with distance-cubed from the local origin. By tracking targets directly in geodetic coordinates, the proposed filter is able to achieve more consistent performance, and a three-fold reduction in computational cost. A series of numerical simulations and field exercises are performed to ground these claims, and verify the efficacy of the proposed filter.
Highlights The proposed filter enables target tracking directly in geodetic coordinates. State update equations are derived from 1st principles, using spherical trigonometry. Performance is shown to match or exceed that of a comparable extended Kalman filter.
Unscented Kalman filter for long-distance vessel tracking in geodetic coordinates
https://orcid.org/0000-0002-4957-6903
https://orcid.org/0000-0002-4188-9614
Abstract This paper describes a unique instantiation of the unscented Kalman filter (UKF) that is particularly well-suited to long-distance surface vessel tracking. The proposed filter leverages a nonlinear process model to predict the position, speed, and heading of targets directly in geodetic coordinates, obviating the need for intermediate coordinate frame transformations, and enabling the use of a simple linear observer. These features differentiate the proposed filter from prior architectures, most of which require the definition of a planar coordinate system in order to describe the linearized state kinematics in an easily differentiable form. Reliance upon local Earth-tangent frames is acceptable in confined operating regions, but incurs substantial estimation error when the target is far from the local origin. It is shown that this error grows with distance-cubed from the local origin. By tracking targets directly in geodetic coordinates, the proposed filter is able to achieve more consistent performance, and a three-fold reduction in computational cost. A series of numerical simulations and field exercises are performed to ground these claims, and verify the efficacy of the proposed filter.
Highlights The proposed filter enables target tracking directly in geodetic coordinates. State update equations are derived from 1st principles, using spherical trigonometry. Performance is shown to match or exceed that of a comparable extended Kalman filter.
Unscented Kalman filter for long-distance vessel tracking in geodetic coordinates
Cole, Blake (author) / Schamberg, Gabriel (author)
Applied Ocean Research ; 124
2022-05-10
Article (Journal)
Electronic Resource
English
Multiple Vehicle 3D Tracking Using an Unscented Kalman Filter
| British Library Conference Proceedings | 2005
Multiple vehicle 3D tracking using an unscented Kalman
| IEEE | 2005
Acoustic Tomography of the Atmosphere Using Unscented Kalman Filter
| Online Contents | 2014