ESTIMATING BUS TRAVEL TIME SAVINGS FROM QUEUE REENTRY PRIORITY
AND JUMPER LANES by
B.S., Montana Tech of the University of Montana, 1997
A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Master of Science Civil Engineering Program
This thesis for the Master of Science degree by Ryan Archibald has been approved for the Civil Engineering Program by
Wesley Marshall, Chair Bruce Janson
Date: May 13, 2017
Archibald, Ryan (M.S., Civil Engineering)
Estimating Bus Travel Time Savings from Queue Reentry Priority and Jumper Lanes These directed by Professor Bruce Janson
Bus preferential treatments at signalized intersections have been and continue to be implemented by transit agencies to reduce bus travel times. One such treatment are queue jumper lanes where the bus can utilize the right turn lane or dedicated jumper lane at an intersection. This allows the bus to bypass a portion of the vehicle queue to service passengers at either a near side or farside bus stop. In the case of a nearside bus stop, the queue jumper lane would typically be designed with an early green signal so that the bus can jump the queue and not be required to wait for the adjacent vehicles to clear prior to reentering the traffic flow. Two studies were conducted to evaluate the need for an early green indication and the design length of the queue jumper lane. The findings were then used in a computer simulation. The studies performed included a field study of bus reentry and a data evaluation of bus arrival locations in a vehicle queue at seven intersections in Denver, Colorado. The studies showed firstly, that the early green indication is not necessary and secondly, that busses, on average, arrive in less than 50% of the 95th percentile queue length. The subsequent computer simulation showed that, for the simulated arrangement of nearside/off-line bus stops, there is a greater delay savings due to short reentry times rather than a longer queue jumper lane.
The form and content of this abstract are approved. I recommend its publication.
Approved: Bruce Janson
Bruce Janson, PhD, Associate Dean of Civil Engineering, University of Colorado Denver.
Wesley Marshall, PhD, Associate Professor Civil Engineering, University of Colorado Denver
Jonathan Wade, Manager of Service Development Support, RTD
Lacy Bell, Manager of Corridor Planning, RTD
Chris Quinn, Planning Project Manager, RTD
Li-Wei Tung, PhD, Senior Transportation Planner, RTD
Amy Rens, P.E., Senior Engineer, City and County of Denver Public Works Transpiration and Mobility Department.
Paul Dreher, Senior Engineering Associate, City and County of Denver Public Works Transpiration and Mobility Department.
Bart Przybyl, P.E., Traffic Engineer, Apex Design, Denver, Colorado.
Jeff Ream, P.E., Traffic Engineer, Apex Design, Denver, Colorado.
TABLE OF CONTENTS
II. REVIEW 01 LITERATURE.................................................4
Additional Literature Gaps.............................................9
Bus Arrival in Queue..................................................15
Eastbound Traffic Volume.............................................22
Westbound and North/Southbound Crossing Volumes......................23
Other Vissim Inputs..................................................23
Base Conditions/No Bus Stops.........................................25
Bus Bay/No Bus Priority..............................................30
Bus Bay/Bus Priority.................................................30
Bus Bay/Bus Priority/Queue Jumper Lane...............................31
LIST OF TABLES
Table 1. Evans and University Eastbound Bus Reentry Observations 14
Table 2. Bus Queue Arrival Location 20
Table 3. Traffic Camera Counts Evans & Broadway 23
Table 4. Vissim Queues 30
Table 5. Travel Time Comparison 32
LIST OF FIGURES
FIGURE Figure 1. Figure 2. Figure 3. Figure 4. Figure 5. Figure 6. Figure 7. Figure 8. Figure 9. Figure 10 Figure 11
Queue Jumper Lane Arrangements Study Intersection Study Corridor
Bus Queue Arrival Location Distribution Broadway Queue Build-Ups 20 Model Runs Logan Queue Build-Ups 20 Model Runs Pearl Queue Build-Ups 20 Model Runs Downing Queue Build-Ups 20 Model Runs Franklin Queue Build-Ups 20 Model Runs High Queue Build-Ups 20 Model Runs University Queue Build-Ups 20 Model Runs
28 28 29 29
LIST OF ABBREVIATIONS
BRT bus rapid transit
Broadway South Broadway Street
CCD City and County of Denver
Downing South Downing Street
Evans East Evans Avenue
Franklin South Franklin Street
GPS Global Positioning System
HCM Highway Capacity Manual
High South High Street
Logan South Logan Street
Pearl South Pearl Street
RTD Regional Transportation District
TCRP Transit Cooperative Research Program
TSP transit signal priority
University South University Boulevard
v/c volume to capacity ratio
vphpl vehicles per hour per lane
Transit agencies across the United States are investing in bus transit priority projects with the goal decreasing travel time and increasing ridership. The Highway Capacity Manual (HCM 2010) lists several transit preferential treatments that can be implemented to provide an advantage to transit over cars. One of these listed is a queue jump or a short bus lane section (often shared with a right-turn lane), in combination with an advanced green indication for the lane, that allows buses to move past a queue of cars at a signal. According to the TCRP (2013), a queue jump lane is a relatively short lane that is available for buses to bypass queues of general traffic at or prior to a signalized intersection, thus reducing delay to bus passengers. For purposes of this paper, the queue jump will be referred in two parts; the queue jump or priority signal at a signalized intersection and the queue jumper lane or the lane a bus can enter (including right turn lanes) to overtake a queue at a signalized intersection. The TCRP (2013) lists three types of queue jumper lanes. The first, which is the focus of this paper, is used with mixed traffic situation. The second and third types are associated with dedicated bus lanes and are not applicable to this study. The first type listed by the TCRP can be further divided into ones with transit signal priority (TSP) and ones without (Figure 1). TSP is typically an early green for the bus allowing it to jump the queue. The types with TSP are for nearside bus stops when the queue jumper lane is not extended through the intersection. The TSP or early green allows the bus to reenter the through lane without having to merge into the through lane, causing delay. The types without TSP are associated with farside stops and since the queue jumper lane extends
through the intersection the bus can reach the farside stop and has room to merge back into traffic.
Nearside Bus Stop and TSP
Farside Bus Stop and No TSP
Figure 1. Queue Jumper Lane Arrangements (TCRP 2013)
In urban areas, the right of way might not exist to add queue jumper lanes at signalized intersections or perhaps a right turn lane exists but downstream right of way is not present to extend the queue jumper lane. In this study, the limited right of way scenario where a queue jumper can be shared with a right turn lane and the queue jumper lane cannot be extended past the intersection is explored. Specifically, is the bus reentry time delay substantial enough to warrant TSP (i.e. what are the delays if TSP is not provided)? TSP can be difficult to implement with transportation agencies since it could add delay to other vehicles because it is uses up green phase from other vehicle movements at the intersection. This study also focuses on the design length of the queue jumper lane.
The delay savings associated with these arrangements depends on two factors. First, how much delay the bus experiences reentering the queue and second, where the bus
typically arrives in the queue and whether it can enter the queue jumper lane. For purposes of this paper, the first is referred to as bus reentry and the second as queue arrival.
REVIEW OF LITERATURE Bus Reentry
Buses will experience a delay reentering traffic when serving passengers at nearside, off-line station at a signalized intersection. The TCRP (2013) has developed methods for estimating this delay which involves two parameters. First, when stopped at a red signal, the queue service time, or the time that queue in the lane adjacent to the bus needs to clear once the signal is green. This assumes that no vehicles queued in the adjacent lane, once moving, let the bus renter so the queue service time is dependent on the adjacent queue length and highly variable based on driver behavior.
Second, the TCRP (2013) method requires that the gap-in-traffic delay be estimated. This is the delay that the bus experiences while waiting for a suitable gap in traffic to merge once the queue has been serviced. The TCRP (2007) states that the gap-in-traffic delay depends on the adjacent lane traffic volume, and various studies have shown that clearance times can range from 9 to 20 seconds. In another reference, the TCRP (2003) assumes clearance times at 15 seconds for 50% green cycle.
The queue service time and gap-in time are then summed together to determine the reentry delay. The TCRP (2013) states that reentry delay is 0-10 seconds at un-signalized intersections and 0 seconds up to the length of the effective green interval at signalized locations. As the literature shows for the reentry delay, the gap-in time portion is highly variable. The queue service time portion is dependent on the adjacent queue length so is also highly variable. However, this portion of the equation could be less variable based on state bus merge laws or bus driver behavior (i.e. at nearside stops, stopping short of the bus stop to
have more merge length or cheating into the intersection after servicing passengers). The focus of the reentry delay will therefore be on the queue service time and determining whether it is a more constant value.
Further literature, regarding the queue service time, was researched to determine if the actual time has been estimated. If the TCRP method is used by agencies to model existing conditions, then the bus delay will be greater in the model than in actual conditions. Delay savings from bus route enhancements through preferential treatments, like transit signal priority, could then be overestimated since the existing condition model is not accurate. This could lead to investment in preferential treatments that do not result in the delay savings anticipated and hoped for mode shift to transit. This is important since just minutes of delay savings are important as evident with the TCRP (2003) which theoretically estimates a five-minute bus running time savings affects mode shift. Further study conducted by Currie and Sarvi (2012) using a quantitative basis, showed that mode shift from automobile drivers at levels of travel savings much lower than theoretical models.
The queue service time is applicable to nearside stops and studies of bus delays at nearside stops are numerous but few address the queue service delay. Furth and SanClemente (2006) used a deterministic model to evaluate the net delays for near and farside locations since a case for time savings can be made with both types. They determined that when a bus has an exclusive lane, however, placing the stop at the stop line yields negative net delay, making it (slightly) superior to far-side placement. The bus-only lane is more applicable to full bus rapid transit and the evaluation did not include off-line stops where queues must be served. Computer simulations have also been conducted to determine delays. Shrestha et. al. (2010) used CORSIM simulations to determine optimal nearside
location but again, no off-line stops were considered. Zhou and Gan (2005) used Vissim traffic modeling software with three queue jumper design options. The first being a queue jumper without transit signal priority or a similar situation to the intersections studied in this paper. They found a delay in lighter traffic of over 15 seconds and in heavier traffic, or 0.95 volume to capacity (v/c) ratio of over 30 seconds. They were not explicit as to whether the queue was serviced but it can be assumed that they were since the values represent multiple car headways.
The TCRP (1996) estimated queue jumper savings time to buses in the field. To do this, the researchers compared the bus travel time to the travel time of a vehicle entering the system at the same time. The intersection had a farside bus stop. Six buses were measured during high volume periods resulting in an average travel time savings of 6.5 seconds. However, the study did not include nearside bus stops where the bus can take advantage of red light time to service passengers.
Farid et, al. (2015) conducted an evaluation of queue jumper lanes performance in reducing travel time using both an analytical (equation based) and simulation approach. Performance was reported in the following person-based measures; person delay and person discharge. The equations used in their study could be relevant to this research since they can be used to estimate transit delay at nearside bus stops with queue jumper lanes. For the queue jumper evaluation, they evaluated two cases, the bus unable to enter the jumper lane upon arrival due to traffic queues extending longer than the jumper lane and second, queues short enough to allow the bus to enter the jumper lane. In both cases, they assumed that the queue jumper lane included a downstream merge lane. The queue service delay is not included in the analytical models since the bus would have lane to accelerate and merge.
Therefore, the research does not recreate the situation in many urban intersections where a right turn lane that could be used as a queue jumper but no merge lane downstream since the right of way might be utilized for street parking.
Zlatkovic et. al (2014) studied an actual bus rapid transit (BRT) system in Vissim. The base condition included buses in mixed traffic, nearside stops, with no queue jumper lanes. Buses could therefore block traffic at nearside stops. The queue jump simulation included an eight second leading green and a 240-foot nearside queue jumper lane and 90-150 foot farside bus bays. Queue Jumpers resulted in a BRT travel time reduction of 15.6% in the westbound direction and 15% eastbound. Lahon (2011) performed a similar simulation and found a 30% reduction in travel time with queue jumper lane and transit signal priority. Both studies dealt with farside stops and lack of literature could be found regarding use of a right turn lane. The apparent focus of research is on farside stops since the consensus is that the farside alignment provides more delay reduction compared to the nearside arrangement even though Gu et. al. (2014) stated that the matter of nearside and farside bust stop location in relation to time savings is yet to be settled.
Further analysis with computer simulation only was conducted by Fitzpatrick and Nowlin (1997). The purpose of the study was to determine the optimal bus stop design: curbside, open bay (off-line), or queue jumper. The modeling program they used was calibrated with field data through code adjustments by the software developers. The simulation for the queue jumper lane included one stop with a 300-foot upstream queue jumper lane and 250 foot downstream to a farside bus stop. The results showed a 5 to 33 second bus travel time savings for volumes below 1000 vehicles per hour per lane (vphpl). Greater reductions were charted above 1000 vphpl. The focus of this study was again,
farside bus stops. Buggs et. al. (2016) conducted a microscopic evaluation but only assumed jumper lanes and farside stops. The savings for the farside was 1-5 %.
The research identified focuses largely on queue jumper lanes with farside bus stops. Less research was identified regarding nearside stops utilizing an existing right turn lane and no downstream merge without TSP
Where the bus arrives in relation to the queue of vehicles is important to the design of queue jumper lane. The longer the queue jumper lane, the better the chances that the bus is not blocked from entering the lane if the signal is red and vehicles have queued. With nearside stops as studied in this paper, the bus is able to use the red phase to service passengers. The TCRP (2014) states that random bus arrivals should be assumed. Xue-song (2014), while studying bus arrival and the ability to predict, assumed that at intersections,
the vehicles reaching a crossroads and participating in a queue approximately obey the Poisson. Bie et. al. (2012) studied arrival time at signalized intersection using global positioning system (GPS) but assumed no nearside bus stops and did not track where the bus arrived in a relation to the queue.
Additional research on the bus arrival location in the queue was not identified, likely since arrivals have been shown to follow the Poisson distribution. In addition, field collection of the data or traffic camera data analysis is likely difficult to achieve. Field data for a bus arrival location in relation to the intersection stop bar but not in relation to the queue is available. Comparing this data to simulated queue lengths is the approach of this study.
Additional Literature Gaps
As discussed, the focus of this study is queue service time and if TSP is required for nearside stops and where does the bus arrive in a queue to optimize the queue jumper lane length. Research was not identified addressing the determination of bus arrival locations from field data and two computer simulation aspects discussed below; determining the distribution of queue lengths in Vissim and simulating bus early entry without a priority signal.
Two delay causes for buses entering and exiting off-line stops or bus bays at signalized intersections include the amount of time necessary for a bus to reenter traffic once passengers are served and the ability of the bus to enter the bus bay or queue jumper lane (i.e. is the vehicle queue due to the signalized intersection blocking entrance in to the que jumper lane and bus stop). Both delays can be considered when designing bus queue jumper lanes. Two separate studies were conducted to help estimate bus time savings due to the addition of nearside queue jumper lanes at signalized intersections. The first study focused on queue service time and whether the queue is serviced at a signalized intersection prior to the bus merging at an off-line stop. If the queue is not serviced, then what is the number of vehicles that proceed through the intersection prior to the bus merging. The results from this study were used in the Vissim model discussed in subsequent sections.
The second study focused on determining bus arrival location in the queue in relation to the stop bar to help determine optimal queue jumper lane lengths. Actual bus positioning data was used for this study.
The objective of this study is to determine if, at a red phase in a signalized intersection with nearside, off-line bus stop, the vehicle queue must clear or is serviced prior to the bus reentering traffic once the signal turns green. The study was conducted at an urban arterial intersection in Denver, Colorado. Specifically, the intersection of East Evans Avenue (Evans) and South University Boulevard (University). The Denver Regional Transportation Districts (RTDs) bus Route 21 travels along Evans through the University intersection both
east and westbound. The eastbound route has a nearside, off-line bus stop that is part of the
right turn lane. Also, the intersection is two major arterials with heavy traffic allowing ample situations where the bus would be required to reenter traffic after servicing passengers.
Figure 2. Study Intersection
Colorado currently has a yield-to-bus law passed in 2009 which essentially makes it possible that the queue is not serviced assuming motorists obey the law. According to the Denver Post (2011), In 2009, Colorado legislators passed the Yield to Bus Law to help transit agencies that were finding that the inability of buses to get quickly back into the traffic flow after a stop was hurting their on-time performance.. The Colorado State Patrol Drivers Handbook (2014) states:
Drivers in the same lane of traffic behind a transit bus (such as an RTD bus, for example) are required to yield the right-of-way to the bus if the bus, after stopping to allow passengers to board or exit, is signaling to enter a traffic lane and the YIELD warning sign on the back of the bus is illuminated. These yield signs are a warning to drivers behind transit buses that they are required to yield when the bus is entering a traffic lane.
They study is therefore confirming compliance with this law and if verified, confirming that the queue service delay in the TCRP Report 165 is not applicable. Methodology
Data was collected by conducting visual observations of the intersection during arrival of the eastbound RTD Route 21. Only the eastbound route was studied since eastbound University has the bus stop in the right turn lane. Observations were taken at peak times when there was sufficient traffic for a vehicle queue to form during the red phase in the through traffic lane adjacent to the bus stop. Peak time also ensured that there was enough passenger servicing so the dwell time, for the most part, used up the remaining green cycle. This limited the buses from servicing then reentering on the same green as arrived (if it arrived on a green). The Route 21 headways are 30 minutes during peak travel times so observations were recorded over several days. Observations noted included:
Bus arrival time.
Signal phase active when the bus arrived.
Whether cycle failure, or missed green phase, occurred.
If cycle failure occurred, which signal phase did the bus arrive on the second time.
If the bus arrived on the green phase, whether the bus departed on the same green after servicing passengers.
Bus location in the queue if not able to enter the right turn lane and bus stop.
Where the bus would have been in the queue if not entered the right turn lane or bus stop.
Once the phase was green, was the queue serviced or not.
Once the phase was green, how many vehicles proceeded through the intersection prior to the bus reentering.
Reentry time with the start being when the light turned green and ending when the bus was fully in the adjacent lane.
Though much was observed, the objective of the study was met with collection of the number of vehicle that proceeded through the intersection prior to the bus reentering once the signal was green. The remaining observations were collected to verify modeling date if necessary.
The average number of vehicles that proceeded through the intersection prior to reentry of the bus was 2.6. This is based on 16 measurements collected over 14 separate days.
Table 1. Evans and University Eastbound Bus Reentry Observations
Date Arrival Time Vehicles Until Reentry
10/21/2016 8:40 AM 1
10/21/2016 4:37 PM 2
10/25/2016 4:32 PM 4
10/26/2016 4:25 PM 0
10/26/2016 4:39 PM 2
10/27/2016 4:37 PM 2
10/28/2016 3:47 PM 4
10/30/2016 4:37 PM 2
11/1/2016 4:36 PM 2
11/2/2016 4:39 PM 2
11/3/2016 4:38 PM 9
11/4/2016 4:08 PM 1
11/7/2016 4:42 PM 2
11/8/2016 4:44 PM 4
11/10/2016 4:40 PM 2
11/15/2016 4:10 PM 2
Standard Deviation 1.968
Confidence in the sample size was determined with the following equation from Hayter (2012):
n > 4 x ( --------)
\ L0 /
n = 14.9 or 15 samples required (16 taken)
n = sample size needed to reach specified confidence level (oc) of 95% tK/2,n-i= critical point of sample size (given) = 1.960 S = standard deviation of sample set = 1.968 (see Table 1)
L0 = confidence interval length (1 desired) = 2
Therefore, there is a 95% confidence interval that the average vehicles prior to reentry within 1 of 2.6. The number is not closer to 0 since, typically, two vehicles were queued adjacent to the bus and could not allow reentry without pausing once the signal was green and allowing the bus ahead. In addition, the immediately adjacent vehicles cannot see the flashing yield light on the bus. The average of 2.6 indicates that the motorists are obeying the law and allowing the bus to merge since the first two cars would likely not see the flashing merge signal.
This reentry delay of 2.6 vehicles is used in the modeling discussed below and indicates that the queue is not serviced prior to bus reentry. This leads to model more closely aligned to actual field conditions.
Bus Arrival in Queue
If the typical bus arrival location in a queue at a signalized intersection is known, the optimal queue jumper lane length can be determined to allow the bus to enter the lane without being blocked. This could result in time savings on the order of a signal cycle time. For example, if the queue jumper lane was not extended, the bus may have to wait for an entire red phase then proceed to the lane and bus stop then service passengers and likely miss the green phase. A queue jumper lane could aid the bus to arrive during the initial red, load during the red, then proceed at green. To determine the bus arrival location, RTD data was evaluated to determine the average distance from the stop bar the bus arrives. Data was collected for University but also expanded to the west in the Evans corridor to allow for a more intersections to be included in the computer simulation.
The corridor studied contains intersections along Evans in Denver, Colorado, including the University intersection included in the reentry study. There are seven signalized intersection circled in Figure 3 below; two major/major intersections (South Broadway Street [Broadway] and University) and 5 major/minor with Evans (South Logan Street [Logan], South Pearl Street [Pearl], South Downing Street [Downing], South Franklin Street [Franklin], and South High Street [High]). Of these intersections, only Broadway and University have right turn lanes. For the Broadway intersection, the bus stop is a farside configuration so the bus does not use the right turn lane. In the case of University, as discussed, the bus stop is nearside so the right turn lane is in a sense, utilizing a queue jumper
RTD (2017) operates over 125 bus routes in the Denver metropolitan area. RTD uses GPS receivers on buses to gather real-time data. One peripheral component of the GPS receivers is a gyroscope that among other things senses when the bus stops. According to the provider init (2013) when a stop is sensed, position data is sent from the GPS receiver to the data recorder on the bus. Additionally, when the bus is stopped, the device records whether the door is opened. Stops when the bus door is not opened are therefore stops in traffic such as in the queue of a signalized intersection. From this data, the location from the stop bar can be determined.
To obtain the data from the recorders on the Route 21 buses, latitude and longitude for the Evans arterial from Broadway to University was submitted to RTD and data within this area for Route 21 was requested. A year (November 2015 to November 2016) of bulk data from RTD was supplied representing 1,468,278 individual records. The following methodology was used to determine the location the buses stopped in the queue.
1. Bulk data in text form was imported into a database.
2. Queries for each intersection along Evans were created.
3. Queries included the following fields based on information provided by init (2013):
a. External IDs: RTD assigned number specific to the bus route. Route 21 departing from their origin on weekdays between 3:05 PM to 5:42 PM were selected. These routes arrive within the study arterial at PM peak.
c. Scheduled start time.
d. Doors Opening: Data collected includes whether the bus door is open. This is used to discern the difference between a stop in traffic versus a stop at a bus stop.
e. Stop Type: Data collected include the type of stop such as a scheduled stop or a disturbance stop (stop in traffic or at a light). This was used as a cross check to the doors opening field.
f. Latitude of the location that the bus stopped.
g. Longitude of the location that the bus stopped filtered to capture only those stops associated with the intersection.
h. Arrival time or stop time (seconds converted to time of day on 24-hour clock).
4. The area defined by the longitude at each intersection analyzed as well as the coordinates of the stop bar was determined using itouchmap.com which is a user interface utilizing Google Maps data. Typically, the area from the stop bar east to the next intersection was used.
5. Queries were run and the data was exported to a spreadsheet.
6. The distance from the stop bar to the stopped location of the bus was determined using a conversion from latitude/longitude to feet. The conversion equation is available from Dunn (2017) found on the Office of Surface Mining Reclamation and Enforcement website.
7. Equivalent car lengths of this distance were determined assuming a 25-foot space used by vehicles.
8. Data was filtered in the spreadsheet eliminating stops where the door was opened (actual bus stops and not queue stops as desired) and any other stops not related to the traffic signal, such as crosswalks and upstream bus stops.
9. For the Broadway intersection, further filtering required to eliminate the bus stopping events caused by a stop sign before the merge onto Evans from the frontage road.
Also, eastbound buses backtrack to reach the Evans Station. This backtrack requires a stop to turn left which was recorded by the data recorders. After the Evans Station, they proceed to the merge mentioned above.
10. Combined data from each intersection into one table and evaluated outliers which led to three data records dismissed for the Logan intersection.
11. Built distribution charts to check for a Poisson distribution. Evaluating the distribution also helped identify outlier data. Distributions are found below.
S. Broadway St. S. Logan St.
Figure 4. Bus Queue Arrival Location Distribution
Note: University lias an existing nearside bus stop in the right turn lane which serves as a queue jumper lane. Therefore, there are no queue arrival locations recorded that are less than the jumper lane length.
The University distribution shows a lack of data under 10 feet vehicle lengths. This is because the buses can enter the right turn lane when the queue is less than 10 cars and proceed to the nearside bus stop to load. These stops are not recorded as queue since the
doors are opened at the stop. For this reason, the University RTD data was not further evaluated since the average value could not be determined due to this data skew.
From this conglomeration of data, the length from the stop bar was averaged by time
period for each intersection. Results are found in the following table:
Table 2. Bus Queue Arrival Location___________________________
_________Location from Stop Bar (feet)________
Intersection______4:00 to 5:00 PM 5:00 to 6:00 PM Average
Broadway 193 191 192
Logan 212 222 217
Pearl 136 103 119
Downing 248 240 244
Franklin 100 108 104
High 143 165 154
University NA NA NA
Note: Values for University not determined.
These lengths are used in determining the queue jumper lanes lengths to be included in the model.
Now that the reentry time and average bus stop location are known, a model was developed to determine a means of evaluating queue jumper length and delay savings in relation to the 95th percentile vehicle queue length at a signalized intersection. The 95th percentile queue length was chosen since this is a number typically available to transportation agencies from signal modeling software.
The modeling was conducted with Vissim which uses the microscopic approach to describe traffic flow. According to Gruber et. al. (2015), the microscopic approach sometimes referred to as the car-following theory or the follow-the-leader theory, considers spacings between consecutive vehicles and speeds of individual vehicles. Conversely according to Gruber et. al. (2015) in a macroscopic simulation the simulation is sectioned off and individual are not tracked. The microscopic approach was selected for this evaluation since it allows for simulation of the nearside/off-line bus stop where the bus is required to reenter after 2.6 cars and can leave the normal traffic flow and enter the queue jumper lane.
Traffic volume data was obtained from the City and County of Denver (CCD) Public Works, Transportation and Mobility Division. Specifically, the volumes used by CCD for their Evans corridor simulations for signal cycle length and coordination design (using Synchro software). To perform an eastbound simulation more closely associated with the same time period as the RTD queue data, recent eastbound traffic volumes from 4:00 PM to 6:00 PM from CCD Broadway traffic cameras were used. Vissim requires volume input at the beginning of the link, or Broadway in this case. Therefore, the CCD traffic camera
volumes will be used for the east end of the link. For the westbound and cross street volumes, the CCD simulation volumes will be used.
Eastbound Traffic Volume
Bulk traffic camera data (79,525 individual records) for eastbound Evans at the Broadway intersection were obtained from CCD. This represents nearly four months of data from September 27, 2016 to January 11, 2017. To obtain the traffic volume entering the corridor at Broadway from 4:00 PM to 6:00 PM, the following methodology was used:
1. Imported the text data from CCD to a database (time column attribute required a change the change to short integer to facilitate import).
2. Created and ran a query table to filter the volume counts to obtain the 4:00 PM to 6:00 PM measurements only.
3. Exported the table data to a spreadsheet and deleted blank or zero volume counts since these were likely a camera error.
4. Added a day of week calculation in since only weekday volumes were desirable.
5. Filter data to obtain the 4:00 PM to 6:00 PM data and determined the hourly volumes based on all movements.
6. The volumes of the 4:00 PM to 5:00 PM and the 5:00 PM to 6:00 PM were averaged resulting in a volume of 1,372 vehicles per hour for eastbound Evans (Table 3).
Table 3. Traffic Camera Counts Evans & Broadway
Time Period Average of 15 Minute Volumes Hourly Flow
4:00 to 4:15 PM 333
4:15 to 4:30 PM 339
4:30 to 4:45 PM 361
4:45 to 5:00 PM 362 1394
5:00 to 5:15 PM 363
5:15 to 5:30 PM 332
5:30 to 5:45 PM 330
5:45 to 6:00 PM 324 1349
Westbound and North/Southbound Crossing Volumes.
Volumes for westbound Evans and north and southbound Broadway, Logan, Pearl, Downing, Franklin, High, and University were obtained from the CCD simulation data. The routing decisions, or turn volumes, for each intersection were also obtained from the CCD simulation data.
Other Vissim Inputs
Signal timing and phasing was obtained from CCD traffic signal plans implemented in 2015 and 2016.
Signal coordination was obtained from the CCD simulation.
Lane orientation, turn bay lengths, lane width, lane numbers, and speed limits were derived from Google Maps.
Detector data was not used but rather maximum green times for all directions were inputted at each intersection to simulate saturated conditions.
Bus headways of 15 minutes were used which does not match the actual RTD Route 21 headway of 30 minutes but allowed for more data collection
Using the input data above, the following Vissim models runs were conducted for listed purpose:
Base Conditions/No Bus Stops: Obtain 95th percentile queue length to use for design of the queue jumper lanes.
Nearside Bus Bays/No Bus Reentry Priority: Modeled at the seven signalized intersections in the corridor assuming the queue must be serviced prior to reentry when stopped at red. Bus stop locations do not match existing conditions but were modeled in this configuration to meet the intent of this evaluation.
Nearside Bus Bays/Bus Reentry Priority: Modeled at the seven signalized intersections in the corridor assuming the reentry priority studied above.
Nearside Bus Bays/Bus Reentry Priority/Queue Jumper Lanes: Modeled at the seven signalized intersections in the corridor. Queue jumper lane length is based on the 95th percentile queue length and bus arrival locations as discussed below.
Nearside Bus Bay/Bus Only: Modeled to isolate the travel time savings based on queue savings. This isolated the dwell time and the control delay since they will be constant.
The models were simulated 20 times for each condition beginning at seed 42 with 1 seed increments. This provided a stabilization of the data. Vissim generated error logs which was reviewed for the runs and fixed as necessary and re-simulated.
Base Conditions/No Bus Stops
The first simulation was used to estimate the 95th percentile queue length to be used for the queue jumper lane lengths. The following process was used to arrive at the 95th percentile queue length from Vissim.
1. Queue counters were added in Vissim at each signalized intersection.
2. Queue length data from 1,800 to 5,400 was collected every 5 seconds. Beginning collection after 1,800 seconds allows the model to seed.
3. After 20 runs, the queue length data for the 7 intersections was copied into a spreadsheet.
4. Deleted lengths less than one car length (25 ft).
5. Filtered the data by queue counter number or intersection. Vissim reports a queue length and a queue max length. The queue length measurement is the actual simulated que length during the time interval. The queue max length is a calculated or estimated value. To obtain the resulting simulated queue length and not the estimated, a shorter measurement timeframe of five seconds was used. Therefore, to determine the actual maximum length the peak value from each red phase buildup, the five second measurements had to be isolated. This was accomplished with a calculation within the spreadsheet. The queue builds and peak values per intersection are found in Figures 5 to 11.
6. Once the peak queue length for each red phase buildup was determined, the distribution was plotted and the 95th percentile interpolated (Table 4).
The base conditions model simulation queue lengths and signal coordination were field checked at 4:00 PM to 4:30 PM on February 16, 2017 and 5:00 PM to 5:30 PM on
Queue Length (feet)
February 17, 2017. This led to vehicle speed adjustments since the portion from Broadway to Logan is an uphill grade. The resulting Vissim 95th percentile queue lengths are as follows:
U") o o c~t o r~-00 00 ON
rt co so 00 CO O',
i/~) O IT) lO o
rf j M O O t ^ O t'
rirtrJMMMrtiN 00*n000<^)0 o
M M M M M ft M M
O O /"> O O (S O V) V)
so r- oo oo cs o
M M N CJ M fTj
IQ V) O IQ O IQ
fQ; ^ O O IQ Tt
so r- oo oo on o
rt rt ct n cs (Q
..... so r- oo oo os o
n00">0*/")/'T0t/rt0t/")0^00t/rt0 rriOOO'O^-OOTtfQO'MfMOOO Dctri|iQiQOr'OOixO'Oi-ir)rt(Q
't 'T 'It '
c^i c^i c^i co '
i r^i m cn co m
o m o iy^; o o tn o o o to
Model Time Perios (seconds)
Figure 5. Broadway Queue Build-Ups 20 Model Runs
o '> o o
00 00 Os o
r~- so 00 CO ON O
(N M (S M
O O m 3 U') iA}
Model Time Perios (seconds)
Figure 6. S. Logan Queue Build-Ups 20 Model Runs
OOOiM^t^COOO-Ht O-HCSC^, ^-w^tnOOOOsOst_J^
in^oooiOHTfinOMOOOOHN^ioo -* ' ininiOh00 00OOHHN(fi
. . _ _ O -H M
vi /-) v>
hOOHM^i^hOOOOt'ii-inooo CSC^TfA)*^)'Or^'G50sO cs>
-^in'ONhhO-^in^'OOOHNin'Or'OHN V')*/")SDr-000005 ^^tt^^00000\0 M fC
Model Time Perios (seconds)
Figure 7. S. Pearl Queue Build-Ups 20 Model Runs
O00O'3-f0^-a500r-^-r0o _. rHrHr-NN(S(S(S(SNNcsNcsNmmfn^iMfnww(ncncn OO'OOO'OOOU^OW^O'OOU^^OOV^O
o co -t r^i o Oi co o I- rj n o m h o 7t n ^ a
^j O O O 05 O O M N ^ -t t
^- -^J- ^f- -^f- y^ y^j y^i
Model Time Perios (seconds)
Figure 8. S. Downing Queue Build-Ups 20 Model Runs
^inininoinoin^^ino^oo^i OiHO5rj-(SC0M0\O^tiri'O^a>05 coo\oo^MMtn't^lo^ooo\050
M 0\ O O h N N fO 1"0 'O h 00 C\ a O m >i
_ /-.l r^\ rf' i^\ r rV5 rV5 r5\ _i
coo)-Hioffir'r'Ooa5f^^,^->o'tco --- ----- -- t-h
Model Time Perios (seconds)
Figure 9. S. Franklin Queue Build-Ups 20 Model Runs
O O 0
<0 Cs 00 Cs Os
ui */i ui */i o
Os t-- t|-
ONMW^1 (N (N N (N N
O O O O /1
Os t- Tf C^i
ONMr.i-M (N N M (S
0 0 0*00
*/S SO SO t 00 CS
*0*0*0 0*0 Os On 00 ^J- m -t *0 O h M
roj On c-
Os On O ...... .
00 00 <0
<0 <0 <0 <0 <0 <0 Tt
<0 <0 <0 <0 <0 <0 Tf
Model Time Perios (seconds)
Figure 10. S. High Queue Build-Ups 20 Model Runs
<Â§ 1000 ob
*0*0*0 O CO so
CO CO os . ..... . _ ......
6 6 6'
O 00 so
00 00 Os _ ...._ ....
*o ^^^^^^*0 0 0 0 0
Model Time Perios (seconds)
Figure 11. S. University Queue Build-Ups 20 Model Runs
The following table lists the 95th percentile queue lengths for each intersection. This data is then compared to the RTD bus arrival locations to determine the average location in
the queue that the buses arrive. This average is then used to determine the design queue
jumper lane length.
Table 4. Vissim Queues
Interpolated 95th RTD Arrival Design QBL
Percentile Queue Locations Length based on
Intersection Length (feet) (feet) % of 95th Average
Broadway 444 192 43% 205
Logan 331 217 66% 153
Pearl 282 119 42% 131
Downing 546 244 45% 253
Franklin 240 104 43% 111
High 404 154 38% 187
University 1125 NA NA 520
Percentages are below 50% except for the Logan intersection where the existing conditions have a downstream farside stop at the Broadway intersection. Busses would serve passenger at Broadway then end up further back in the Logan queue.
Bus Bay/No Bus Priority
As previously discussed, the base condition model was modified to include nearside off-line bus stops or bus bays at the seven intersections in the study. Priority to the buses was not modeled so if the phase was red when passenger servicing is complete a queue likely forms in the lane adjacent to the bus. This model simulates the queue clearing or servicing before the bus can renter once the light turns green. Like the base conditions run, 20 simulations were conducted. Bus travel times after the 1,800 second seed period were tracked and averaged. Results are found in Table 5.
Bus Bay/Bus Priority
The bus bay model was further modified to allow for bus priority. A priority rule was added in the model at each of the bus bays. The priority parameters (oncoming vehicle headway and oncoming vehicle speed) were adjusted until the bus could enter the traffic prior to queue serving. Exact reentry at two to three cars was difficult to achieve using
Vissim priority rules so the reentry at each intersection from one to six at Broadway and University and zero to two at Logan, Pearl, Downing, Franklin, and High. Since Vissim is a microscopic model, the accelerations and merge times vary from driver to driver so reentry varies. Regardless, the model simulates that queue not servicing as intended. If a bus serviced passengers on, priority was not given and the bus was required to gap-in. This is similar to field conditions since vehicles at speed would not stop to let the bus in nor would the bus have the yield light active. Twenty simulations were conducted and bus travel times after the 1,800 second seed period were tracked and averaged. Results are found in Table 5. Bus Bay/Bus Priority/Queue Jumper Lane
For this simulation, the bus bays were extended with a nearside bus queue jumper lane keeping the priority rule. The length of the queue jumper lanes modeled was determined based on the average RTD bus stopped location versus the 95th percentile calculated in Vissim. Twenty simulations were conducted and bus travel times after the 1,800 second seed period were tracked and averaged. Results are found in Table 5.
Finally, a bus only (no other vehicles in the system) simulation was conducted to isolate the delay savings strictly due to queuing. Buses experience delay from traffic controls, dwell time, and queueing. The first two delays are constant within the model so conducting a bus only can help determine the savings due to the improvements.
Model results are summarized in the following table.
Table 5. Travel Time Comparison
Model Run Travel Time (seconds) Delay Savings Queue Time (seconds) Queue Savings
No Cars 542 NA NA NA
Bus Bay/No Bus Priority 816 NA 274 NA
Bus Bay/Bus Priority 672 18% 130 52%
Bus Bay/Bus Priority/Bus Bypass Lane 655 20% 113 59%
Compared to the RTD Route 21 posted run time of 10 minutes from the Evans Station to University, the Bus Bay/No Bus Priority travel time of over 13 minutes. However, the travel times are not directly comparable since the modeled travel time is based on seven nearside stops where the queue must be serviced. The actual bus route has a mix of nearside and farside bus stops, both on and off-line.
The results of this study indicate that for bus route priority treatments 1) an early green that allows the bus to queue jump may not be necessary, especially in a state with yield-to-bus laws, 2) for corridors like the one modeled, bus priority reentry and queue jumper lanes modeled to approximate 50% of the modeled 95th percentile vehicle queue length at signalized intersections save 18% and 20% travel time, respectively. The bulk of the delay savings are from the reentry priority and not the queue jumper lane. This indicates that in a nearside arrangement along the corridor, queue jumper lanes need only be as long as the right turn lanes and lengthening them does not add to significant delay savings. From module simulation reviews, this is likely because since the buses had priority, they could reach the next intersection towards the front of the platoon and longer queue jumper lanes had no real advantage. The arrangement in this corridor of nearside stops allows the bus to take advantage of dwell time during the red phase.
The bus travel time savings should be treated as estimates only since traffic corridors vary significantly. Furthermore, the modeled 95th vehicle queue length was based on no bus trips in the corridor where buses might influence the length due to on-line stops or slower speeds. Since the Route 21 headways are 30 minutes, it is assumed that they would have little influence on the 95th percentile queue length.
Further evaluation should be conducted to create a model of the existing conditions. This would include modeling of the actual bus stop arrangements and a study of the current RTD data to determine how often passengers are served and the boardings and alightings at bus stops in the corridor. This could provide a travel time evaluation from existing to
improvements. Also, a model for farside off-line and farside on-line arrangements could be created to compare differences in travel time for near versus farside. Nearside arrangements were evaluated since a right turn lane may already be in place and be used for queue jumper lanes. Farside right-of-way might be used for parking and removing parking is typically an unpopular decision and difficult to achieve for transit agencies. Finally, a sensitivity analysis could be conducted to determine the delay savings associated with various queue bypass lane lengths.
Bie, Y., Whang, D.,and Qi H. (2012). Prediction Model of Bus Arrival Time at Signalized Intersection Using GOS Data. Journal of Transportation Engineering, 138(1), 12-20.
Bugg, Z. Crisafi, J., Lindstrom, E., and Ryus, P. (2016). Effect of Transit Preferential Treatments on Vehicle Travel Time. Proc., 95 th Annual Meeting of the Transportation 37 Research Board, Washington D.C.
Colorado State Patrol. (2014). Colorado Driver Handbook.
Currie, G., and Majid, S. (2012). New Model for Secondary Benefits of Transit Priority. Transportation Research Record: Journal of the Transportation Research Board, 2276, 63-71.
Dunn, M. (2017). Latitude Longitude Distance Calculator.
(Jan. 14, 2017).
Farid, Y.Z. Christofa, E., and Collura, J. (2015). Dedicated Bus and Queue Jumper Lanes and Signalized Intersections with Nearside Bus Stops. Transportation Research Record: Journal of the Transportation Research Board, 2482, 182-192.
Fitzpatrick, K., and Nowlin, R. L. (1997). Effects of Bus Stop Design on Suburban Arterial Operations. Transportation Research Record: Journal of the Transportation Research Board, 1571, 31-41.
Furth, P.G., and SanClemente, J.L. (2006). Nearside, Farside, Uphill, Downhill. Transportation Research Record: Journal of the Transportation Research Board, 1971, 66-73.
Garber, N.J., and Hoel, L.A. (2015). Traffic & Highway Engineering, 5th Edition. Cengage Learning, Stamford, CT.
Gu W., Gayah, V., Cassidy, M., and Saade, N. (2014). On the impacts of bus stops near signalized intersections: Models of car and bus delays. Transportation Research, B(68), 123-140.
Hayter, A. (2012). Probability and Statistics for Engineers and Scientists, 4th Edition, Brooks/Cole Publishing Co., Boston, MA.
Haufe, P. (2013). MOBILEreports Manual, init, Chesapeake, VA.
Lahon, D. (2011). Modeling Transit Signal Priority and Queue Jumpers for BRT. ITE Journal, December 2011, 81(12), 21-24.
Leib, J. (2011). RTD buses getting rear-facing yield lights as part of new law. (Jan. 7, 2017).
Regional Transportation District (2017). Schedules. (Feb. 1, 2017).
Shrestha, P.K., Nakamura, F., and Okamura, T. (2010). Assessing the Optimal Location of Nearside Bus Stop to Enhance Advantages of Bus Signal Priority (BSP). Journal of the Eastern Asia Society for Transportation Studies, 8, 1314-1328.
Transit Cooperative Research Program (TCRP) (1996). Report 19 Guidelines for the Location and Design of Bus Stops. Transportation Research Board, Washington D.C.
TCRP (2003). Report 90 Bus Rapid Transit Volume 2: Implementation Guide, Transportation Research Board, Washington D.C.
TCRP (2007). Report 118 Bus Rapid Transit Practitioners Guide, Transportation Research Board, Washington D.C.
TCRP (2013). Report 165 Transit Capacity and Quality of Service Manual, 3rd Edition, Transportation Research Board, Washington D.C.
Transportation Research Board (2010). Highway Capacity Manual. Transportation Research Board, Washington, D.C.
Xue-song, C. (2014). Collaborative prediction for bus arrival time based on CPS. Journal of Central South Liniversity, 21, 1242-1248.
Zhou, G., and Gan A. (2005). Performance of Transit Signal Priority with Queue Jumper Lanes. Transportation Research Record: Journal of the Transportation Research Board, 1925, 265-271.
Zlatkovic, M., Stevanovic, A., and Zahid Reza, R.M. (2013). Effects of Queue Jumpers and Transit Signal Priority on Bus Rapid Transit. Proc., Transportation Research Board 2013 Annual Meeting, TRB Paper # 13-0483, Washington D.C.