1. 1 Copyright © 2015 by ASME
Proceedings of the 2015 Joint Rail Conference
JRC2015
March 23-26, 2015, San Jose, CA, USA
JRC2015-5774
A METHOD TO VERIFY RAILROAD INTERCONNECT WITH HIGHWAY TRAFFIC
SIGNAL SYSTEMS
Adam Moore
City of Portland
Portland, Oregon, USA
Paul Zebell
City of Portland
Portland, Oregon, USA
Peter Koonce
City of Portland
Portland, Oregon, USA
Jon Meusch
Northwest Signal
Portland, Oregon, USA
ABSTRACT
In response to increasing concern about railroad grade
crossing safety, the Federal Railroad Administration and
Department of Transportation issued Safety Advisory 2010-02
recommending in part “...that railroads conduct comprehensive
joint inspections of highway traffic signal pre-emption
interconnection with State and local highway authorities…”
2010-02 also recommends recording devices at interconnected
highway-rail grade crossings.
This paper addresses a method to facilitate these goals by
enabling the highway authority to independently verify that rail
equipment is functioning properly, and just as importantly,
enabling the railroad to independently verify that the highway
traffic signal equipment is providing adequate clearance time in
advance of the arrival of the train in the crossing. The method
involves adding two circuits between the rail equipment and the
traffic signal equipment: a crossing island circuit, and a start of
the traffic clearance phase indicator from the traffic signal to the
rail equipment. This system has been implemented at two
intersections in Portland, Oregon, with plans for further
implementation.
INTRODUCTION
Signalized intersections located within 200 feet (61 m) of
rail grade crossings require a traffic signal preemption protocol
to allow highway traffic to exit the rail right-of-way in the event
of a train passing. By alerting a traffic controller of a train’s
approach, the system can change the sequence of the traffic
signal to allow traffic to promptly exit the crossing and prevent
vehicle entrapment. Additionally, the preemption can prevent
conflicting traffic control messages during the train’s passing
that would otherwise direct traffic through the crossing during a
train’s approach or occupation.
Rail crossings that see high traffic volumes or that have
safety concerns are outfitted with active warning devices:
flashing lights with or without gates. When a rail crossing with
active warning devices is located near a signalized intersection,
these two control systems are combined and referred to as an
interconnected crossing. Interconnected crossings are considered
the highest form of safety treatment at an intersection and
railroad crossing short of grade separation or closure [1].
Despite the interconnection of railroad and highway
systems, accidents still occur at these crossings. Car-train crash
events are significantly more likely to result in fatalities or
injuries than other types of intersection crashes [2]. Following
prominent coverage of the collision between a school bus and a
commuter train in Fox River Grove, Illinois, in 1995, few efforts
have been made at the local, state, and national level to improve
interconnected crossing safety.
BACKGROUND
In 2010, the Federal Railroad Administration (FRA) issued
Safety Advisory 2010-02 in response to Safety
Recommendations I-96-10 and I-96-11 issued by the National
Transportation Safety Board (NTSB). SA 2010-02 recommends
that states, local highway authorities, and railroads install,
maintain, and upgrade railroad and highway traffic signal
recording devices at interconnected crossings. The safety
advisory also recommends joint inspections of traffic signal pre-
emption interconnections between highway and railroad
authorities, and the use of recording device data during these
inspections to aid in the evaluation process.

2. 2 Copyright © 2015 by ASME
SA 2010-02 recommends the use of railroad and highway
traffic signal recording device data at interconnected crossings
to provide a record of any anomalies associated with the
interconnected system operation. The safety advisory notes that
“recording devices should be capable of recording sufficient
parameters to allow railroad and highway personnel to readily
determine that the highway traffic signals and railroad-activated
warning systems are coordinated and operating properly.”
Modern traffic controllers already record a variety of system
functions at signal-controlled intersections. Event logs provide a
wealth of information about an intersection’s functionality and
allow authorities to efficiently address performance shortfalls.
Similarly, railroad active warning systems record the status of
railroad warning devices. What has historically been missing is
a unified data log of railroad and traffic signal device operations.
Interconnected datasets can assist with interconnected
crossing diagnostics in ways that separate data and/or field
observations cannot. Preemption designs are dependent on
consistent cooperation between traffic signal systems and
railroad active warning systems, and interconnected datasets
provide archived performance that can reveal trends and
anomalies. Trends in interconnected crossing performance allow
for the visual display of “real-world” operating conditions that
may differ from the preemption design. Anomalies that may be
missed during field observations are logged and easily
discernable.
Following the issuance of SA 2010-02, the need for an
interconnected recording solution has become more prominent.
While supplemental recording devices capable of receiving data
from traffic signal systems and railroad active warning systems
are available for purchase, it is more economical and practical
for agencies to incorporate rail active warning data in existing
traffic controller equipment. This strategy meets the
recommendation for data collection put forth by the FRAwithout
the need for additional equipment, installation, or training.
This paper proposes a method to facilitate highway-railroad
system interconnected datasets using existing traffic controllers
with slight modifications to the cabinet relay configuration.
PROPOSED INTERCONNECT METHOD
An impending train arrival initiates a sequence of warning
events and alarms at an interconnected crossing:
1. An “outer” circuit allows time to clear the longest conflict
with the track clearance phase (e.g. ped clearance).
2. An “inner” circuit provides ample warning time to
complete any vehicle phases that conflict with the track
clearance phase.
3. Vehicle clearance phase starts.
4. Railroad warning devices are activated (e.g. gates, lights,
and/or bells).
5. The train occupies the crossing.
6. The preemption sequence ends.
Typically, highway traffic-related events (e.g., advance and
simultaneous preemption) are recorded by the traffic controller
and railroad active warning-related events (e.g., gates down) are
recorded by the railroad active warning system.
The integration of an island circuit indicating the presence
of a train in the railroad crossing enables the traffic controller to
record the time that the crossing is occupied by a rail vehicle.
For the application at the interconnected crossing in this
study, the interface with the railroad consists of three mechanical
double-pole double-throw general purpose 120-volt relays, two
DC isolators, and one AC isolator (for the island circuit). The
sequence of preemption events is as follows:
1. A rail vehicle traveling on the approach towards the
crossing above a minimum speed (nominally 5 mph)
activates a railroad outer circuit.
2. The railroad constant warning devices activate an inner
circuit when the estimated time of arrival is less than the
warning time required by the traffic signal.
3. The gate and other warning devices are set in motion.
4. The gate down circuit is activated when the gates are
horizontal.
5. The island circuit is activated when a train is detected in
the crossing.
6. The preemption sequence ends.
By utilizing an island circuit, the traffic signal system is able
to record all railroad active warning system activity (along with
train presence) without the need for additional equipment while
meeting the recommendations for recording railroad and traffic
signal activity put forth by the FRA SA 2010-02.
CASE STUDY
The proposed interconnect method detailed in the previous
section has been implemented at two intersections in Portland,
Oregon. The interconnected crossing at SE 11th
Avenue and SE
Gideon Street is one such crossing.
SE 11th
Avenue is classified as a collector street and sees
significant traffic volumes, particularly during peak hours. The
speed limit is 30 mph [48.2 km/h]. Freight, transit, and school
buses utilize the corridor. The intersection with SE Gideon Street
is approximately 300 feet (90 m) downstream from a Union
Pacific railroad crossing outfitted with an active warning system
that includes flashing lights and gates. Signalized pedestrian
crossings are provided at the SE Gideon Street intersection and
approximately 150 feet (45 m) upstream at the SE Clinton St
intersection. These crossings are 0.6 miles (1 km) from the Union
Pacific Brooklyn rail yard and see upwards of 25 Union Pacific
train crossings and several more Amtrak train crossings a day.
Because the rail corridor includes Amtrak trains travelling at
higher speeds, the approach is exceptionally long.
An aerial view of the interconnected crossing is provided in
Figure 1, and Figure 2 shows the layout and phase orientation
that highlights the unusually complex operations of the
intersection. The intersection complexity will increase in

3. 3 Copyright © 2015 by ASME
September 2015 when a new LRT line comes online. This
addition to the intersection is reflected in the Figure 2 diagram.
The traffic controller at the intersection is a Type 2070
running Voyage software, which records railroad active warning
events. These data are available for download through network
communications with the city central system in conjunction with
the intersection signal data. Recorded as “alarms,” the start and
end of each railroad-related event is noted with a time stamp. The
first four alarms are received from the railroad active warning
system. Alarm 1 is the outer circuit pedestrian clearance
notification. Alarm 2 is the inner circuit signal to the traffic
controller to begin a vehicle track clearance phase. Alarm 3 is a
signal that the active warning gates are horizontal. Alarm 4
indicates the train is present in the crossing, made possible by the
integration of the island circuit. Alarm 5 is generated by the
traffic controller, indicating the green track clearance phase is
underway. The start (“on”) and end (“off”) of each alarm is
recorded. For each train crossing, ten alarms should be recorded.
Alarm 1 (on and off): Outer circuit
Alarm 2 (on and off): Inner circuit
Alarm 3 (on and off): Gates horizontal
Alarm 4 (on and off): Train present
Alarm 5 (on and off): Green track clearance
The design values for the interconnected crossing are ideally
as follows. The outer circuit (Alarm 1) activates at 67 seconds
prior to train arrival. The inner circuit (Alarm 2) activates at 40
seconds prior to train arrival. Green track clearance (Alarm 5)
allows a minimum of 31 seconds to clear the tracks. However,
because of the variability of the long approach, the actual times
are much greater.
Data Description
Data for this paper were pulled from October 7-25, 2014,
consisting of about 5,500 alarm entries. In its raw format, the
proposed interconnect method dataset requires a degree of post-
processing for analysis. To best judge performance, events
(alarms) are inspected relative to the presence of a train, i.e.
Alarm 4. This gives useful information such as time between
initiation of the green clearance interval and the train arrival. A
simple script can calculate times between Alarm 4 and the other
alarms for each train crossing.
Reflecting the nature of “real-world” operations, alarm
sequences are sometimes incomplete or inconsistent. This is
likely the result of unusual train activity due to the proximity to
the rail yard that sees, for example, trains partly entering the
crossing before reversing or two trains crossing in opposite
directions at the same time
For the purposes of this paper, incomplete alarm sequences
were not included in the analysis. An incomplete alarm sequence
is an alarm sequence during a train crossing that does not include
a warning event (e.g., outer circuit is missing, or the gates are not
recorded as lowering) or includes too many warning events (i.e.
the train entered the crossing, departed, and then arrived again
without the gates raising). Incomplete sequences do not indicate
a faulty interconnected crossing, but they do make the analyses
in this study – in which alarms are measured from one instance
of Alarm 4 per train crossing – complicated and confusing.
An inconsistent alarm sequence occurs when alarms are
initiated out of order. Many inconsistent alarm sequences were
observed in the study dataset.
Data post-processing involved removal of incomplete alarm
sequences and calculation of alarms relative to the presence of
each train crossing. Inconsistent alarm sequences were not
removed. The resulting dataset is representative of typical
activity at the interconnected crossing.
Analysis
For the purposes of this study, the analysis consisted of
summary and visual data representations that demonstrate the
ability of recorded data to describe interconnected crossing
activity in simple and intuitive ways.
After post-processing, descriptive statistics were calculated
for every train crossing. The primary concern in this study was
using active warning system data to calculate alarm times
relative to the arrival and departure of a train; thus, values are
given relative to the start of Alarm 4, when the train had just
arrived, or relative to the end of Alarm 4, when the train had just
departed. Statistics are presented in Table 1 (train arriving), Table
2 (train just arrived), and Table 3 (train departed).
Figure 1. Aerial view of interconnected crossing. Railroad crossing at
top of image; SE 11th Ave/SE Gideon St intersection at bottom of image.
(Image: Google Maps. North towards top of page.)

4. 4 Copyright © 2015 by ASME
Figure 2. Intersection layout and phase orientation. Note future LRT line
to be put in service in September 2015.
Table 1. Summary Statistics for the Study Dataset for Each Railroad
Active Warning Event Prior to the Arrival of the Train
Times Relative to Start of “Alarm 4: Train Present”
Alarm 1
On
to
Alarm 4
On
Alarm 2
On
to
Alarm 4
On
Alarm 5
On
to
Alarm 4
On
Alarm 3
On
to
Alarm 4
On
Description
Outer
Circuit
Inner
Circuit
Green
Track
Clearance
Gate
Horizontal
Design
Value
(sec)
67 40 31 18
Min (sec) 11 11 4 -23
Max (sec) 162 148 142 124
Mean (sec) 77 70 63 28
Median
(sec)
77 70 63 26
St. Dev.
(sec)
19 20 19 16
Minimum, maximum, and mean times help give an
understanding of the extremes and typical operation of the
interconnected crossing. The pedestrian clear-out interval circuit
activates, on average, 77 seconds before the train’s arrival. The
green track clearance begins 14 seconds after activation of the
outer circuit at 63 seconds prior to the train’s arrival. The green
track clearance lasts until eight seconds after the train’s arrival.
Table 2. Summary Statistics for the Study Dataset for Each Railroad
Active Warning Event Just After the Arrival of the Train
Times Relative to Start of
“Alarm 4: Train Present”
Alarm 4 on
to
Alarm 5 off
Description
Green Track
Clearance
Minimum (sec) -91
Maximum (sec) 56
Mean (sec) 8
Median (sec) 7
St. Deviation
(sec)
13
Table 3. Summary Statistics for the Study Dataset for each Railroad
Active Warning Event After the Train has Departed
Times Relative to End of “Alarm 4: Train Present”
Alarm 4 Off
to
Alarm 1 Off
Alarm 4 Off
to
Alarm 2 Off
Alarm 4 Off
to
Alarm 3 Off
Description
Outer
Circuit
Inner Circuit
Gate
Horizontal
Min (sec) -45 0 4
Max (sec) 47 47 51
Mean (sec) -0.4 2 6
Median (sec) 1 1 5
St. Dev.
(sec)
9 4 5
Table 4. Summary Statistics for the Study Dataset for Length of Time
Train was in Crossing
Alarm 4 On
to
Alarm 4 Off
Description Train Present
Min (sec) 11
Max (sec) 2629
St. Dev. (sec) 333
Mean (sec) 232
Median (sec) 121

5. 5 Copyright © 2015 by ASME
Negative values indicate an alarm occurring out of typical
sequence. For instance, the minimum value of the Gate
Horizontal alarm in Table 1 is -23 seconds. This value indicates
the train arrived in the crossing before the gates were lowered.
Considering the nearby rail yard, such a situation can occur
during rail switching movements, when a train approaches the
crossing at a very slow speed, stops before entering the crossing,
and then finally enters the crossing, at which point the gates are
lowered.
The data also provide insight into the train activity.
Specifically, the length of time the train was in the crossing is
simple to discern. Table 4 details the time trains spent in the
crossing, ranging from 11 seconds to three quarters of an hour.
Data visualizations provide an intuitive understanding of the
interconnected crossing activity. Figure 3 depicts the standard
deviations of alarms relative to train arrival (Alarm 4 On) at the
crossing. The standard deviation of each alarm from Table 1 and
Table 2 is plotted on a timeline. One and two standard deviations
are provided from the average initiation of each alarm. Data
points that are outside two standard deviations are covered by
railroad operating rules.
Figure 4 shows a box plot distribution of alarms prior to and
just after a train’s arrival at the crossing. Box plots are convenient
ways of depicting variation in data, showing the median (thick
black line), the quartiles (box sides), maximum and minimum
(whiskers), and outliers (dots). Figure 4 is useful in identifying
inconsistent alarms, such as the gate horizontal alarm (Alarm 3
On) activating after the arrival of the train (Alarm 4 On – fixed
at t = 0 seconds); such behavior can be attributed to the crossing’s
proximity to the Brooklyn rail yard, in which a train may come
within inches of the island circuit as it makes switching
movements. In such an instance, the record will indeed show the
train in the intersection prior to the gate being horizontal. In a
similar manner, the outer circuit can come on after the inner
circuit if a train stops on the approach and then starts toward the
crossing.
Figure 5 utilizes the same box plot concept to show the
variation in train occupation at the crossing. Figure 5 reinforces
the very large standard deviation in train occupancy noted in
Table 4, and the data are clearly skewed. Extremely long dwell
times can likely be attributed to the nearby rail yard and typically
occur late at night.
While box plots are useful in visualizing data variation, they
do not reveal any assumptions about the distribution of the data.
Figure 6 shows a probability density function of alarms prior to
and just after a train’s arrival at the crossing. A probability
density function displays the probabilities of alarm activations
and deactivations (relative to Alarm 4 On or Alarm 4 Off in this
context). The sharper a spike, the more likely an alarm is to occur
at that time. A wider distribution indicates a less predictable
occurrence of an alarm. For instance, Alarm 5 Off (end of green
clearance) consistently occurs after the arrival of a train, while
the initiation of the outer circuit (Alarm 1 On) occurs over a
wider range of times.

6. 6 Copyright © 2015 by ASME

7. 7 Copyright © 2015 by ASME
DISCUSSION
The summaries and visualizations of interconnected
crossing data in this paper show the power of archived data to
better understand a crossing’s performance and to identify
anomalies. The results show that standard traffic controller
equipment with slight modifications can adequately record
crossing data from both railroad active warning systems and
highway traffic control systems, meeting the recommendations
for recording crossing activity put forth by the FRA Safety
Advisory 2010-02.
Benefits associated with archiving and subsequently
visualizing interconnected crossing data are numerous. Figure 3,
Figure 4, Figure 5, and Figure 6 intuitively provide information
on the crossing’s performance. One can easily see that the alarms
are generally occurring in compliance with the crossing order
design, in the correct sequence, and at the desired intervals,
though with a degree of variability. It follows that the
intersection and railroad crossing are effectively working
together. In many ways, these figures offer much more value than
a typical field observation, as they allow a look at the
interconnected crossing activity over an extended period of time.
In the absence of the interconnected crossing performance
analysis presented in this paper, the value of recording data is
still substantial in that authorities will have a timeline of
interconnected crossing activity in the event of any incidents or
complaints.
In the same way that the data allow a highway or railroad
agency to determine that the interconnected crossing is
functioning properly, the data also highlight shortcomings in the
crossing’s performance. Providing ample preemption time for
the pedestrian clear-out interval is crucial in the overall ability of
the intersection to clear vehicles prior to the train’s arrival; in the
sample dataset, the outer circuit alarm activates as early as 162
seconds and as late as 11 seconds before the train arrives. Each
of these extremes can be considered excessive and in need of
attention.
There are two issues with the interconnect method proposed
in this paper. First is the time required to outfit an intersection
with the slight modifications, though it should be noted that any
effort to record interconnected crossing data requires some
degree of installation and configuration time. The Voyage

8. 8 Copyright © 2015 by ASME
software used in this case study may be unique in its
functionality. The second issue is the need for data post-
processing. While it is ultimately left to individual highway and
railroad authorities to decide data quality tolerances, it is the
opinion of these authors that some data will be filtered out to ease
an analysis. On the aggregate level, this means abnormal train
activity, as was observed at this interconnected crossing due to
the nearby rail yard, will be excluded from statistical summaries
and visualizations. Excluded data will need to be examined on a
case-by-case basis for whatever it can reveal, perhaps
lengthening the interconnected crossing evaluation process.
Collection and analysis of interconnected crossing data will
encourage and simplify efforts of states, local highway
authorities, and railroads to conduct periodic joint inspections of
crossing performance by allowing independent examination at
each agency’s discretion prior to collaboration. This benefit is
especially paramount in areas that have a large number of
interconnected crossings or, as in this case, crossings that see a
large number of train crossings per day that would benefit greatly
from an aggregate analysis over a large time frame. Automating
the data collection and reporting process will further enhance the
ability of agencies to quickly inspect interconnected crossings.
CONCLUSIONS
FRA Safety Advisory 2010-02 recommended
comprehensive joint inspections and implementation of
recording devices at interconnected crossings. This paper
outlines a method for utilizing traffic controllers with slight
modifications for collection of both traffic signal system and
railroad advance warning system data, meeting the FRA
recommendation.
A data set is used to demonstrate the ability of the proposed
method to evaluate and visualize interconnected crossing
performance. Advantages of aggregated data are discussed that
include summaries and visual presentations of data, allowing
identification of anomalies and poor performance.
The process used in this study to examine the interconnected
crossing is unique, due in large part to the lack of documentation
and guidelines set forth in state and federal guidelines. The
authors note an explicit need for more understanding and
guidance in interconnected crossing analysis to facilitate
consistency across the country.
ACKNOWLEDGEMENTS
The authors wish to thank Tom Urbanik and David Stubbs
for their assistance with this study.
REFERENCES
[1] Institute of Transportation Engineers. 2006, “Preemption of
Traffic Signals Near Railroad Crossings,” Washington, DC.
[2] Datta, T. K., Gates, T. J., Savolainen, P., Fawaz, A., and
Chaudhry, A., 2013, “Timing Issues for Traffic Signals
Interconnected with Highway-Railroad Grade Crossings,”
OR10-31, Detroit, MI.