Analysis of the neural extended Kalman filter for target tracking using different neural network functions

Authors:
Stephen C. Stubberud;Kathleen A. Kramer
Affiliations:
Rockwell Collins, Inc., Poway, CA;University of San Diego, San Diego, CA
Venue:
AIA '08 Proceedings of the 26th IASTED International Conference on Artificial Intelligence and Applications
Year:
2008

Citing 5
Cited 0

System identification: theory for the user

System identification: theory for the user
Training multilayer perceptrons with the extended Kalman algorithm

Advances in neural information processing systems 1
Artificial neural systems: foundations, paradigms, applications, and implementations

Artificial neural systems: foundations, paradigms, applications, and implementations
Bayesian Multiple Target Tracking

Bayesian Multiple Target Tracking
Predictive guidance intercept using the neural extended Kalman filter tracker

Control and Intelligent Systems

Quantified Score

Hi-index	0.00

Visualization

Abstract

The neural extended Kalman filter is an adaptive estimation technique that has been shown to improve target-tracking performance when the target is maneuvering. The technique relies upon a neural network which is trained on-line to modify the target motion model. Different mathematical functions have been proposed and implemented as the hidden-layer squashing function of the neural network. For a general tracking application where a wide variety of targets with different maneuver specifications are present, the performance of these different hidden layer functions is analyzed to provide a baseline metric for meaningful comparison and evaluation. Using these results, the neural extended Kalman filter tracking system with the overall best tracking performance for manoeuvring targets can be selected.