1-5-6 
that satisfied this tolerance were considered error free for the 
purpose of mapping the track and the antenna positions were 
averaged. 
Figure 5. Comparison of the computed and measured baseline 
length 
Figure 6 is a map of the track compiled from the four days data 
after all the epochs containing errors had been removed. In total, 
approximately 95% of the track between Ipswich and Forest Hill 
was mapped using the kinematic GPS receivers. 
Figure 6. Map of the track determined using two kinematic GPS 
receivers on the train (arrows indicate missing sections) 
To complete the map of the track, a total station (Sokkia Set 4B) 
was used to undertake a terrestrial survey throughout the largest 
gap in the kinematic GPS data set. The survey was conducted by 
traversing between reference points placed by rapid static GPS 
techniques. Connecting the terrestrial survey to points of known 
coordinates enabled the terrestrial and kinematic GPS surveys to 
be transformed into the same datum. A complete coordinate file 
of the track (consisting of approximately 1200 points) was then 
created (Figure 7). 
Figure 7. Map of the track created by combining the GPS and 
terrestrial surveys (the box indicates the location of the terrestrial 
survey) 
Sensors and Test Setup 
Once the coordinate file of the track was created for map 
matching purposes, a second test was designed to analyse an 
integrated DGPS, dead reckoning and map matching system. It 
was intended that the system would simulate a real-time system, 
but operate in the post-processing environment. To achieve this, 
precise time tagging of multi-sensor data capture was essential if 
an accurate analysis was to be undertaken. Since the platform is 
in motion, inaccurately time stamped data directly translates into 
positional errors. Therefore, the timing system must be able to 
provide time stamps to an accuracy higher than the required 
accuracy of the entire system. The time stamp enables the data 
from each component to be linked together and post-processed in 
a real-time scenario. 
The data sources within the test included a GPS receiver and a 
tachometer. Although it was intended that measurements from a 
gyroscope would also be utilised in the integration, time stamping 
three different data sources was cumbersome. Since the GPS 
receiver could provide three-dimensional position, heading and 
velocity measurements, it was considered that it could suitably 
replace the gyroscope for this preliminary investigation. A 
computer was utilized to log two communication ports and time 
stamp the measurements. Since, a time stamp was only required 
to provide relative time for the two measurement sources, the 
CPU clock was able to satisfy this requirement. Any small 
relative error (< 0.1 s) in the timing of the measurements will 
have an insignificant effect on the resultant position given that the 
acceleration of a train is so small (~1 m/s 2 ), and therefore the 
velocity is unlikely to change significantly. However, in other 
applications, where the vehicle is more dynamic, timing errors of 
this magnitude would be inadequate. 
Table 1 contains the summary of the measurements used and their 
respective sources. 
Measurement Source 
Measurement 
OmniSTAR 3000LR8 
(Trimble - Lassen Sk8 
GPS receiver with 
OmniSTAR differential 
correction service) 
Latitude (<|>) 
Longitude (X) 
Height (h) 
Heading (a) 
Velocity (v GPS ) 
Train Tachometer 
Velocity (v T ) 
Table 1. Measurements and their respective sources 
Once captured, the data was filtered and map matching 
techniques were applied in an attempt to determine on which 
track the train was travelling. The results from the map matched 
positions were then tabulated for further analysis. 
7 RESULTS AND DISCUSSION 
Figure 8 to Figure 10 present the east, north and horizontal 
position residuals between the filtered positions and five parallel 
tracks.