A new source difference artificial neural network for enhanced positioning accuracy
- 1 August 2012
- journal article
- research article
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 23 (10)
- https://doi.org/10.1088/0957-0233/23/10/105101
Abstract
Integrated inertial navigation system (INS) and global positioning system (GPS) units provide reliable navigation solution compared to standalone INS or GPS. Traditional Kalman filter-based INS/GPS integration schemes have several inadequacies related to sensor error model and immunity to noise. Alternatively, multi-layer perceptron (MLP) neural networks with three layers have been implemented to improve the position accuracy of the integrated system. However, MLP neural networks show poor accuracy for low-cost INS because of the large inherent sensor errors. For the first time the paper demonstrates the use of knowledge-based source difference artificial neural network (SDANN) to improve navigation performance of low-cost sensor, with or without external aiding sources. Unlike the conventional MLP or artificial neural networks (ANN), the structure of SDANN consists of two MLP neural networks called the coarse model and the difference model. The coarse model learns the input-output data relationship whereas the difference model adds knowledge to the system and fine-tunes the coarse model output by learning the associated training or estimation error. Our proposed SDANN model illustrated a significant improvement in navigation accuracy of up to 81% over conventional MLP. The results demonstrate that the proposed SDANN method is effective for GPS/INS integration schemes using low-cost inertial sensors, with and without GPS.Keywords
This publication has 14 references indexed in Scilit:
- Knowledge-based neural network approaches for modeling and estimating radon concentrationsEnvironmental Progress & Sustainable Energy, 2012
- The Utilization of Artificial Neural Networks for Multisensor System Integration in Navigation and Positioning InstrumentsIEEE Transactions on Instrumentation and Measurement, 2006
- Bridging GPS outages using neural network estimates of INS position and velocity errorsMeasurement Science and Technology, 2006
- Particle filters for positioning, navigation, and trackingIEEE Transactions on Signal Processing, 2002
- A tutorial on particle filters for online nonlinear/non-Gaussian Bayesian trackingIEEE Transactions on Signal Processing, 2002
- Training the random neural network using quasi-Newton methodsEuropean Journal of Operational Research, 2000
- A new method for the nonlinear transformation of means and covariances in filters and estimatorsIEEE Transactions on Automatic Control, 2000
- EM-ANN models for microstrip vias and interconnects in dataset circuitsIEEE Transactions on Microwave Theory and Techniques, 1996
- Novel approach to nonlinear/non-Gaussian Bayesian state estimationIEE Proceedings F Radar and Signal Processing, 1993
- High-Accuracy Kinematic Positioning by GPS-INSNAVIGATION: Journal of the Institute of Navigation, 1988