arrival times at highway-railroad grade crossing
DESCRIPTION
Estimating Train Arrival Times at Highway-Railroad Grade Crossing Using Multiple Sensors. Presented at the 89th TRB Annual Meeting – Jan 2010.TRANSCRIPT
By:Diego B. Franca, M.Sc.
Kittelson and Associates, Inc
Elizabeth “Libby” Jones, PhDAssoc. Professor, Civil Engineering, University of Nebraska -
LincolnAssoc. Director , Mid-America Transportation Center
Presented at 89th TRB Annual Meeting – Jan 2010
Estimating Train Arrival Times at Highway-Railroad Grade Crossing Using Multiple Sensors
Presentation OutlinePresentation Outline
IntroductionProblem StatementLiterature ReviewMethodologyData CollectionData Analysis and ResultsFinal Considerations
IntroductionIntroduction
Highway-Railroad Grade Crossings (HRGCs) Overview
“The general area where a highway and a railroad’s right of way cross at the same level, within which are included the railroad tracks, highway, and traffic control devices for highway traffic traversing that area” (MUTCD, 2007)
IntroductionIntroduction
HRGCs Concerns– Collisions at HRGCs represent the second largest
cause of fatalities in railroads (trespassing is the first).
– HRGCs near signalized intersections present a safety concern due to the potential queues of highway traffic that can back up across the tracks.
IntroductionIntroduction
Traffic Signal Preemption at HRGCs– The MUTCD requires a 20-second minimum time for
the railroad circuit to activate warning devices prior to arrival of a through train.
– When highway-rail grade crossings are within 200 feet of a signalized intersection, preemption should be considered at that location.
Problem StatementProblem Statement
Motivation– The minimum warning time of 20 seconds to the
signalized intersection is not always adequate to safely clear stopped vehicles from the HRGC area.
– The current traffic signal preemption strategies are viewed as a reactive action to trains approaching a nearby HRGC.
– Advance notice of train arrival times at HRGCs can improve signal preemption strategies and reduce accidents.
Problem StatementProblem Statement
General Hypothesis:– The hypothesis of this research is that second
generation technologies can provide accurate advanced notice of train arrival at an HRGC near a signalized intersection of at least a cycle length prior to the arrival of a train at an HRGC.
Problem StatementProblem Statement
Research Objective 1:– Define how the variability of speed measurements
from two 2nd generation technology sensors affects the train arrival time estimation at an HRGC in a multi-track environment.
Research Objective 2:– Determine how to best use the multiple sensor
speed data to improve estimations of train arrival time at an HRGC.
Literature ReviewLiterature Review
Train Detection Technologies:– First Generation: AC-DC, Motion-Sensitive, Constant
Warning Time.
– Second Generation: Doppler Radar, Video Detection, Infrared.
– Third Generation: GPS, Transponders, Positive Train Control.
Literature ReviewLiterature Review
• NCHRP 271 - Traffic Signal Operations near Highway-Rail Grade Crossings (Korve, 1999)
• An Analysis of Low-Cost Active Warning Devices for Highway-Rail Grade Crossings (Roop et.al., 2005)
• Non-Vital Advance Rail Preemption of Signalized Intersections near Highway-Rail Grade Crossings: Technical Report (Ruback, Balke and Engelbrecht, 2007)
• Estimating Train Speeds Using Fused Data from Multiple Speed Detectors (Zhou, 2007)
MethodologyMethodology
1st Objective– Collect train speed data using Doppler radar and video
image detection.
– Manual speeds are needed so the performance of both sensors can be compared.
– Two scenarios considered: multiple trains and single train on the tracks.
Spee
d
Timet1t0
Area = Train Length
MethodologyMethodology
1st Objective– Train arrival time
estimation methodology from the collected train speed data.
– Required steps to estimate train arrival times: train length and train acceleration estimation.
– Actual train arrival time at the studied HRGC needed to compare the estimations from radar and video.
TIME UPDATE (Prediction)
MEASUREMENT UPDATE (Correction)
velocity and acceleration kinematics equations
radar and video speed measurements
Fused speed estimations
Kalman Filter Algorithm
MethodologyMethodology
2nd Objective– Fuse train speed data from
Doppler radar and video image detection.
– Estimate train arrival times using methodology for the 2nd objective.
– Ideally, the fused data should improve the estimates from radar and video detection.
Data Collection SiteData Collection Site
Upstream Salt Creek
Adams St. HRGC
Upstream 44th St.
• UNL HRGC TEST BED
• Installed equipment helps monitor highway traffic.
• Vehicles are constantly caught stopping at the tracks.
Day/Night Camera aimed at the HRGC
Traffic signal near the HRGC
N Adams St.
Camera Location
Camera field of view
NN
N 35th St.
Adams St.
Data Collection SiteAdams Street HRGC LocationAdams Street HRGC Location
Upstream Salt Creek LocationUpstream Salt Creek Location
• Located 1.85 miles upstream of the HRGC location;
• City of Lincoln Public Works installation support;
• Connected to the UNL-ITS Lab via internet.
Data Collection Site
Data CollectionData Collection
1st Objective– Video and radar speed data collected at Salt Creek
location. Single train and multiple-train scenarios considered.
– Manual speeds measured by computing railcar length and time interval on recorded videos.
Data CollectionData Collection
1st Objective– Actual train arrival times
needed to compare estimations from radar and video.
– Camera time stamps from Salt Creek and HRGC locations used to compute actual train arrival times.
Data CollectionData Collection
2nd Objective– Data collected for 1st objectives used.
– Speed data fused using the Kalman filter model and train arrival time estimations obtained from the new speed estimates.
Data Analysis & ResultsData Analysis & Results
1st Objective– Multiple-Train Scenario
RADAR PAIRED T-TEST
T-Statistic P-value95% Confidence Interval for Mean Difference (sec)
1.58 0.14 (-0.61, 3.68)
VIDEO (AUTOSCOPE) PAIRED T-TEST
T-Statistic P-value95% Confidence Interval for Mean Difference (sec)
0.58 0.571 (-2.47, 4.27)
Data Analysis & ResultsData Analysis & Results
1st Objective– Single Train Scenario
RADAR PAIRED T-TEST
T-Statistic P-value95% Confidence Interval for Mean Difference (sec)
3.15 0.007 (0.98, 5.06)
VIDEO (AUTOSCOPE) PAIRED T-TEST
T-Statistic P-value95% Confidence Interval for Mean Difference (sec)
0.09 0.929 (-4.30, 4.69)
Data Analysis & ResultsData Analysis & Results
2nd Objective– Multiple-Train Scenario
KALMAN FILTER PAIRED T-TEST
T-Statistic P-value95% Confidence Interval for Mean Difference (sec)
1.79 0.099 (-0.32, 3.25)
• Kalman filter presented the narrowest 95% C.I.
Data Analysis & ResultsData Analysis & Results
2nd Objective– Single Train Scenario
KALMAN FILTER PAIRED T-TEST
T-Statistic P-value95% Confidence Interval for Mean Difference (sec)
1.44 0.171 (-0.86, 4.44)
• Radar presented the narrowest 95% C.I., but Kalman filter 95% C.I. includes the zero mean difference.
Data Analysis & ResultsData Analysis & ResultsTrain Arrival Time Estimation Comparison
MULTIPLE-TRAIN SCENARIO Count PercentageDifference from Actual Arrival
Time
Radar Estimation
Autoscope Estimation
Kalman Filter Estimation
Radar Estimation
Autoscope Estimation
Kalman Filter Estimation
Within +/- 2 seconds
6 4 8 46% 31% 62%
Within +/- 5 seconds
9 7 12 69% 54% 92%
within +/- 10 seconds
13 13 13 100% 100% 100%
SINGLE TRAIN SCENARIO Count PercentageDifference from Actual Arrival
Time
Radar Estimation
Autoscope Estimation
Kalman Filter Estimation
Radar Estimation
Autoscope Estimation
Kalman Filter Estimation
Within +/- 2 seconds
6 6 7 38% 38% 44%
Within +/- 5 seconds
12 7 13 75% 44% 81%
within +/- 10 seconds
15 12 15 94% 75% 94%
Final RemarksFinal Remarks
Lessons learned from Doppler radar and video performances will help to develop algorithms to predict train arrival times at HRGCs and improve traffic signal preemption (and safety as well).
Data fusion process showed to be reliable and it could improve train arrival time estimations in a real time system application.
In the event of failure of either sensor, the Kalman filter could still be used to provide consistent measurements.
CreditsCredits
Federal Railroad Administration (FRA)
Nebraska Department of Roads (NDOR)
City of Lincoln