476 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 49, NO. 2, MARCH 2000
A Map Matching Approach for Train Positioning
Part II: Application and Experimentation
Samer S. Saab, Senior Member, IEEE
Abstract—In this paper, applications of the map matching
algorithm proposed in [1] are presented. In particular, steady-state
Kalman filters are proposed and applied for filtering the yaw
rate and tachometer signals. In addition, experimental results,
using a quartz yaw rate sensor and axle encoders aboard a freight
train, are included to show the performance of the proposed map
matching algorithm.
Index Terms—Integration of yaw gyro and pulse generator,
Kalman filtering, map matching, train positioning.
I. INTRODUCTION
T
RAIN positioning systems (TPS’s) play a major role in
automatic train control systems. They are used to prevent
undesirable occurrences such as a train-to-train collision, switch
collision, and collision with a fixed buffer. TPS’s are also uti-
lized for automatic train operation and train scheduling. These
applications include speed control, accurate position stopping,
door control, arrival time, and optimization control of the trains
to move a maximum number of passengers or freight. From
the late nineteenth century to the present, TPS have employed
track circuits to identify track occupancy. The track is divided
into track segments each measuring from 100 to 10 000 m with
each segment provided with a transmitter and at least one re-
ceiver. The occupancy of a train extends the length of all track
circuits which contain any part of the train. In the case where
two trains are traveling on the same line and in the same di-
rection, the following train would be prohibited from entering
the last track circuit occupied by the preceding train. Supple-
mentary safe spacing (one or more track circuits) is added to
encompass the following train stopping profile. This discipline
is known as a fixed block system. Such technology results in
large headways, especially for long track circuits. Given the
practical limits on minimum track circuit lengths, fixed block
systems will repeatedly result in below optimal performance
levels. Lately, several railway companies, with support from the
Federal Railway Administration (FTA), are seeking to optimize
train scheduling and reduce, or even disregard, track circuits.
One of the proposed solutions is the introduction of a moving
block occupancy determination. This method assumes that each
train calculates the section of track encompassed by its worst
case stopping profile and combines it with the train's physical
length to determine its occupancy.
The contribution of this paper is the application of the map
matching algorithm proposed in [1] to TPS’s. In addition, fil-
Manuscript received January 1997; revised January 1998.
The author is with the Lebanese American University, Byblos, Lebanon
(e-mail: ssaab@lau.edu.lb).
Publisher Item Identifier S 0018-9545(00)02562-7.
tering procedures for the yaw rate and tachometer signals are
proposed to maximize signal-to-noise ratio in order to obtain
stronger correlation between the map information and the pro-
cessed sensor measurements. The core of this paper presents the
results of an experiment executed on board a freight train using
a quartz yaw rate sensor. These experimental results illustrate
the performance of the algorithm proposed in [1].
In Section II, an overview of various TPS’s is presented. Sec-
tion III proposes filtering processes and an analysis of the yaw
gyro and tachometer signals. In Section IV, the experiment is
described and the results are presented. Concluding remarks are
given in Section V.
II. SELECTIVE TECHNIQUES FOR TPS
Location of the train is determined by the linear distance it
has traveled along the track in the aligned route. For train sched-
uling, operation, and control, this computed profile information
is then transmitted to central control through the wayside com-
munication system. Wheel wear, wheel slip/slide (due to poor
adhesion), and calibration errors will produce an inaccurate rep-
resentation of displacement. Therefore, a tachometer alone is
not sufficient for the TPS. Various TPS proposed by different
companies can be found in [2] and [3]. These proposed posi-
tioning systems utilize different technologies which are based
on either tachometers, global positioning system (GPS), differ-
ential GPS (DGPS), inertial navigation system (INS), Doppler
radar, proximity beacons (electronic signposts), and/or tags (de-
vice located along the track which are used to rezero the train's
actual position). The pros and cons of these systems are sum-
marized in Table I. Reference [4] describes a reduced-order INS
algorithms tightly coupled with a tachometer sensor and a tag
reader or/and GPS for TPS application. The INS consisted of
yaw and pitch rate sensors and a forward accelerometer. Taking
advantage of the insignificant track superelevation and train sus-
pension, the roll rate was modeled as zero-mean noise. Similar
modeling was applied to the train lateral and upward accelera-
tion. The main contribution of this high-priced system was the
compensation of some of the tachometer error due to slip/slide,
exclusively, in the acceleration and deceleration profile. (In Au-
gust 1994, a successful demonstration of this positioning system
was given to the FTA aboard a light rail vehicle in Pittsburgh,
Pennsylvania.)
To replace fixed block technology, tight position accuracy is
necessary. For example, an automated “metro” system must pro-
vide a means for stopping with a typical accuracy of 0.3 m
at stations platforms where normal passenger exchanges take
place. In addition, track identification is required in order to dif-
ferentiate between adjacent lanes. On the other hand, for main-
0018–9545/00$10.00 © 2000 IEEE