Use of Neural Network for the Improvement of Particle Filter Performance in INS/GPS Integrated Navigation System During GPS Signal Outages

A. Lee

Abstract:	Particle Filters have demonstrated better performance over classical Kalman Filters in many applications in recent years. This is especially true in situations where the underlying dynamical system is non-linear and the system noise and the measurement noise in non-Gaussian. The Particle Filter, however, is, generally computational intensive and the availability of measurements is required to update the weighting factors of the particles and sampling of new particles in the state estimation process. If the measurement is not available, then the overall filter performance will be degraded. In this paper, a customized neural network algorithm, based on the existing concept of the Adaptive-Neuro- Fuzzy-Inference-System (ANFIS), is proposed for use in conjunction with the Particle Filtering Algorithm. The propose algorithm is a modified Particle Filter Algorithm augmented with a Fuzzy Inference Engine whose parameters are determined by a neural network. The purpose of the Neural Network is to capture the association of the inputs and outputs of the Particle Filters so that when the complete input set or measurement for the filter is unavailable, the network can still provide the predicted reference measurement and values of other filter parameters. Based on these predicted values, the Particle Filter will continue to function normally. The network operates in two basic modes: the training mode and the prediction mode. Because of the augmented use of the Neural Network that captures the values of the filter parameters, the filter can always run in update mode even when the measurement data is unavailable. To show that a better performance could be achieved with the proposed algorithm than with the basic Particle Filter algorithm, Monte Carlo simulation runs were performed when the measurement data for the algorithm is not available. Applicability of the proposed algorithm for use in the GPS/INS and other Multi-Sensor Data Fusion Systems were also discussed.
Published in:	Proceedings of the 23rd International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS 2010) September 21 - 24, 2010 Oregon Convention Center, Portland, Oregon Portland, OR
Pages:	3090 - 3106
Cite this article:	Lee, A., "Use of Neural Network for the Improvement of Particle Filter Performance in INS/GPS Integrated Navigation System During GPS Signal Outages," Proceedings of the 23rd International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS 2010), Portland, OR, September 2010, pp. 3090-3106.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In