Design and Testing of a GPS Based Track Change Detection System Using Geometry Car Data

H. James Rome and Leonard Allen

Abstract:	This work develops and evaluates decision function software which can determine whether or not the train is traveling on the assigned (correct) track. The real time inputs are 3 ft RMS raw differential GPS, heading rate, and odometer. Fixed inputs are the track maps, and a switch file. It uses a Bayes decision function which outputs the decision and the confidence. The key to success is that trains can only change tracks at switches and there are usually hundreds of GPS measurements between switches. In addition heading rate with odometer provides an independent measure of change at a switch, many measurements map into a single decision. The design was expedited by “Design and Test” procedures based on 4000 miles of AMTRAK Geometry Car data providing data for the algorithms and for verification. A distance based Kalman Filter is utilized to integrate the system and to synchronize real time measurements with the maps. Virtually every form or reasonableness test available is used to reject suspect measurements, and HDOP is factored into the Bayes Decision function. Evaluation of the data as passed through the decision function indicated that false alarm rates are about a factor of 10 less than that which could be expected from a diligent operator. And one would expect that there would be a miss (switch taken but not detected, and not anticipated by an operator), even temporarily, in about one in 1 million switch intervals.
Published in:	Proceedings of the 2007 National Technical Meeting of The Institute of Navigation January 22 - 24, 2007 The Catamaran Resort Hotel San Diego, CA
Pages:	1179 - 1195
Cite this article:	Rome, H. James, Allen, Leonard, "Design and Testing of a GPS Based Track Change Detection System Using Geometry Car Data," Proceedings of the 2007 National Technical Meeting of The Institute of Navigation, San Diego, CA, January 2007, pp. 1179-1195.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In