You are on page 1of 9

Evaluation of Transition Methods of the 170E and

2070 ATC Traffic Controllers after Emergency


Vehicle Preemption
Ilsoo Yun1; Matthew Best2; and Byungkyu “Brian” Park3

Abstract: Modern traffic signal control systems provide emergency vehicle preemption 共EVP兲 capabilities using advanced sensors and
Downloaded from ascelibrary.org by AJOU UNIVERSITY on 03/23/23. Copyright ASCE. For personal use only; all rights reserved.

communication technologies. While EVP strategies have been widely implemented by urban transportation management agencies, few
research efforts have studied the transition methods employed after EVP operations. This paper presents comprehensive evaluations of
EVP transition operations, considering both exit phase control and transition methods under three different traffic volume conditions. The
research employed hardware-in-the-loop simulation, which provides a more realistic evaluation environment because it incorporates actual
signal controllers into the simulation. The evaluation results at a coordinated-actuated signalized intersections in Northern Virginia
indicated that: 共1兲 performance measures varied significantly depending on the EVP transition methods; 共2兲 shortway in the 170E
controller and smooth in the 2070 ATC controller generally outperformed the other transition methods; and 共3兲 the use of exit phases—
available in the 2070 ATC controller—provided significant benefits over the 170E controller.
DOI: 10.1061/共ASCE兲0733-947X共2008兲134:10共423兲
CE Database subject headings: Traffic signals; Traffic management; Simulation; Vehicles.

Introduction EVP call on a cross street, extensive queuing often results on the
mainlines.
Implementing optimal signal timing plans can improve mobility Even though most traffic signal controllers can implement
and safety within urban signalized intersections. In addition to the EVP, they provide slightly different methods for the signal system
timing plans, modern traffic signal controllers are capable of to return to its coordinated state. Traffic engineers would like to
evaluate the performance of various EVP transition methods be-
implementing various control strategies that can improve the ef-
fore implementing them in the field. However, the available traf-
ficiency and safety of urban signalized intersection systems. For
fic simulation models do not include the ability to model vendor-
example, when emergency vehicles 共e.g., fire trucks, ambulances, specific EVP strategies directly. Hardware-in-the-loop simulation
police cars兲 need priority to serve public needs, traffic signal con- 共HILS兲, which combines a microscopic simulation program and
trollers can provide the right of way to these vehicles using signal actual traffic controllers, can effectively evaluate EVP transition
preemption. With the implementation of emergency vehicle pre- methods under varying demand conditions 共Nelson and Bullock
emption 共EVP兲, emergency vehicles can receive green indications 2000兲. Given the uncertainties associated with EVP 共i.e., the time
along their routes to reach their destinations more quickly. One of occurrence, traffic conditions, etc.兲, the process of optimizing
disadvantage of implementing EVP, however, is the cost incurred controller response to EVP is quite complex. Thus, a research
from disrupting the normal phasing and timing plan of the signal effort that identifies the most appropriate EVP transition method
system. This disruption is especially problematic within coordi- within a laboratory environment is imperative to improve mobil-
nated systems where precise coordination is necessary to ensure ity and safety of urban transportation systems.
smooth traffic flow along an arterial or network. In the case of an The purpose of this paper is to explore various EVP transition
methods and identify the method that can best minimize disrup-
1
Research Fellow, Center for Transport Infrastructure Investment, tions of traffic signal operations caused by an EVP occurrence. A
Korea Transport Institute, 2311 Daewha-dong Ilsanseo-gu Goyang-si, test site consisted of four coordinated-actuated signals along Lee
Gyeonggi-do 411-701, Korea. E-mail: ilsooyun@koti.re.kr Jackson Memorial Highway in Chantilly, Va. was developed
2
Engineering Associate, HDR/WHM Transportation Engineering, 504 using a HILS environment with a calibrated VISSIM microscopic
Lavaca St., Suite 1175, Austin, TX 78701-2817. E-mail: matthew. simulation model, four traffic signal controllers, and four control-
best@hdrinc.com ler interface devices. For the study, two types of traffic
3
Assistant Professor, Dept. of Civil and Environmental Engineering, controllers—McCain’s 170E and Econolite’s 2070 ATC—were
Univ. of Virginia, 351 McCormick Rd., Thornton Hall, P.O. Box 400742, used. The EVP transition operations examined in this paper in-
Charlottesville, VA 22904-4742 共corresponding author兲. E-mail: bpark@ clude both the selection of transition methods available in each
virginia.edu
controller and the use of exit phase control under three different
Note. Discussion open until March 1, 2009. Separate discussions must
be submitted for individual papers. The manuscript for this paper was traffic conditions and different locations of EVP calls in the local
submitted for review and possible publication on June 29, 2007; approved cycle timer.
on June 6, 2008. This paper is part of the Journal of Transportation This paper is outlined as follows. The next section briefly ex-
Engineering, Vol. 134, No. 10, October 1, 2008. ©ASCE, ISSN 0733- plains the operation of EVP and prior studies of EVP transition
947X/2008/10-423–431/$25.00. methods. The third section presents the procedure used in this

JOURNAL OF TRANSPORTATION ENGINEERING © ASCE / OCTOBER 2008 / 423

J. Transp. Eng., 2008, 134(10): 423-431


Table 1. Comparison between 170E and 2070 ATC Traffic Controllers
Controllers
Feature 170E 2070 ATC
Transition methods Smooth
Shortway 共1–4 transition cycles兲 Add only
Dwell 共1–4 transition cycles兲 Dwell
共Dwell time: 0 – 255 s兲
Exit phase control No Yes
Minimum requirement control Yes Yes
共Minimum green or pedestrian time兲 共Minimum green or pedestrian time兲
Minimum EVP time control Clearance time Minimum holding time
Maximum EVP time control Up to 255 s Up to 255 s
Downloaded from ascelibrary.org by AJOU UNIVERSITY on 03/23/23. Copyright ASCE. For personal use only; all rights reserved.

study, and the fourth section provides the results of the HILS • Two cycles= 125%;
experiments. The last section includes conclusions and recom- • Three cycles= 117%; and
mendations for future research. • Four cycles= 112%.
The Dwell transition method can only increase the cycle length
up to the maximum cycle length determined by the selected num-
Literature Review ber of transition cycles. Force-off points of all phases are adjusted
by the same percentage change to maintain the same relative
Emergency Vehicle Preemption in 170E and 2070 ATC phase split during transition. In addition, there are also several
Controllers numeric parameters that must be set to initiate and run EVP 共BI
EVP systems are designed to detect approaching emergency ve- Trans Systems 1994兲. These are described later in the paper.
hicles using sensors and provide a green indication to provide the EVP operations and their settings in Econolite’s 2070 ATC
right of way to these vehicles. A brief description of EVP opera- controller are somewhat different from those in McCain’s 170E
tions in the 170E and 2070 ATC traffic controllers, manufactured controller. The 2070 controller provides three transition
by McCain and Econolite, respectively, is explained in this sec- methods—smooth, add only, and dwell. Similar to the 170E’s
tion. McCain’s 170E controller was selected for the study because Shortway, smooth lengthens or shortens the local cycle length to
it is currently being used at the test site, while Econolite’s 2070 return the controller to coordination. After the EVP period ends, if
ATC controller was selected to ascertain the benefits of exit phase the offset value is less than the desired offset value by less than or
selection. equal to 50%, the controller can add up to 20% of cycle length
Once a preemption call from an emergency vehicle is received per cycle. Otherwise, the controller can subtract up to 17% of
in the 170E controller, the controller terminates its current phase
cycle length per cycle to achieve the desired offset value. The
as soon as the minimum requirements 共e.g., minimum green time兲
2070 controller also features the dwell method, which holds co-
have been served and displays green on the subject preemption
ordinated phase green times until the end of a user-specified dwell
phase. Once the preempt signal is removed and the predetermined
clearance time 共i.e., the amount of green time needed for the period 共255 s maximum兲. This dwell period repeats at the end of
emergency vehicle to clear the intersection兲 has elapsed, the con- each local cycle until coordination is regained. The third transi-
troller starts the transition process to return to normal operations. tion method available in the 2070 controller is add only, in which
During the transition process, the controller may display greens the controller can add up to 20% of cycle length per cycle
according to the sequence used during normal operations. In co- 共Econolite Control Products 2004兲.
ordinated systems, the transition period ends after the controller Unlike the 170E controller, the 2070 controller allows users to
achieves the proper offset value—the difference between the mas- specify exit phases that operate after the EVP operation. Up to
ter controller timer and local controller timer—thus returning the two phases may be selected as exit phases, which serve as tran-
controller to coordination 共BI Trans Systems 1994兲. The exact sition phases while the controller attempts to return to normal
EVP and transition operations, however, are vendor-specific. operations. In the 2070 controller, the EVP phases time after the
McCain’s 170E controller provides two transition methods— current phases have served their minimum green times. They con-
shortway and dwell—with between one and four transition cycles. tinue until all of the following have occurred:
The controller achieves the target offset values by adjusting the • Min hold time has elapsed; and
cycle lengths during the transition operation. With the shortway
• The EVP call has been dropped or the max time 共i.e., 255 s兲
transition method, the controller determines the quickest route to
has elapsed.
achieve the proper offset value and increases or decreases the
Here, the min hold time is the minimum number of seconds that
cycle length accordingly within maximum and minimum cycle
length limits. The minimum cycle length is equal to the sum of the preemption phases time, and the max time is the maximum
the phase coordination transition minimums, which may be ad- number of seconds that the hold phases time. There are also sev-
justed by users, while the maximum cycle length is determined by eral numeric parameters 共e.g., minimum preemption green time兲
the number of transition cycles specified by users. The maximum in the 170E controller that must be set to initiate and run EVP
cycle lengths are specified as percentages of the background cycle 共Econolite Control Products 2004兲. The exit phases, if any, then
length as follows 共BI Trans Systems 1994兲: begin to time up to their maximum duration. The differences be-
• One cycle= 150%; tween the 170E and 2070 controllers are summarized in Table 1.

424 / JOURNAL OF TRANSPORTATION ENGINEERING © ASCE / OCTOBER 2008

J. Transp. Eng., 2008, 134(10): 423-431


Prior Studies
Several studies have presented the effects of EVP on traffic signal
systems, specifically the performance of transition methods of
various traffic signal controllers. Obenberger and Collura 共2001兲
reviewed five transition methods—hold/dwell, maximum dwell,
long way/add, short way, and best way/smooth—based on previ-
ous literature to provide a state-of-the-practice assessment of al-
ternative transition methods. They found that no condition
supports the use of any one transition method for EVP. In addi-
tion, they recommended the use of a simulation methodology
using software-in-the-loop 共SILS兲 simulation or HILS to evaluate
more accurately actual transition methods used in the controller
after EVP operations.
Downloaded from ascelibrary.org by AJOU UNIVERSITY on 03/23/23. Copyright ASCE. For personal use only; all rights reserved.

Shelby et al. 共2006兲 compared several transition methods— Fig. 1. Test site: Lee-Jackson Memorial Highway, Chantilly, Va.
dwell, max dwell, add, subtract, shortway, immediate, two-cycle,
three-cycle, etc.—using CORSIM 共TSIS 5.2兲, which was en-
in microscopic traffic simulation models and not on the exact
hanced to support ACS-Lite. Two different sites were studied—an
methods available in traffic controllers. Although the methods
isolated intersection and a four-intersection arterial. The results
available in the simulation models are intended to mimic those in
indicated that the level of saturation is a significant factor in de-
traffic controllers, they do not always do so with complete preci-
termining the optimum transition method. For instance, in the
sion and accuracy. Using HILS, some research efforts have at-
arterial network study, at 80% or less saturation, the best way
tempted to study the exact transition methods used in the field and
method was most effective. At 90% or higher saturation, add out-
are summarized below.
performed best way in some instances. overall, it was found that
Nelson and Bullock conducted a HILS-based study to quantify
the minimax class of transition methods 共i.e., smooth, best way, the impacts of EVP calls and determine the most effective transi-
long/short, shortway兲 outperforms dwell, which can actually de- tion methods 共Nelson and Bullock 2000兲. They used CORSIM
grade performance at low saturation levels. Further, it was found and actual controllers manufactured by Econolite Control Prod-
that a three-cycle transition period performs better than a two- ucts and tested between one and four consecutive EVP calls to
cycle transition period, which in turn outperforms a one-cycle evaluate the performance of the transition methods. The authors
transition period. They also reported that EVP operations ac- found that single EVP calls had little impact on network perfor-
counted for approximately 18% additional delay to the traffic mance while multiple consecutive EVP calls spaced at short in-
network. tervals greatly affected network performance. In the evaluation of
Obenberger and Collura 共2007兲 also investigated four transi- the various transition methods, they found that the smooth
tion methods—best way, long, short, hold—available in the Si- method performs best.
emens NextPhase Suitcase Tester 共Version 1.4B9兲 traffic signal Yun et al. 共2007兲 evaluated various EVP transition methods
controller software, which was coupled with CORSIM 共TSIS 5.2兲 using HILS, which consisted of a well-calibrated VISSIM micro-
to establish a SILS environment. Using a four-coordinated- scopic simulation model, four 170E traffic controllers, and four
actuated traffic signal system along the Columbia Pike corridor in controller interface devices. The transition methods under exami-
Arlington, Va., they used the midday period from 11:00 a.m. to nation were shortway and dwell with different numbers of cycles
1:00 p.m. as a base traffic condition, as well as 20 and 40% for transition completion and two phase sequences for the cross
increases in traffic volumes from the base conditions. They found streets. The evaluation results indicated that:
that statistically significant interactions exist between the transi- 1. A single EVP call causes significant increases in delays and
tion methods and traffic conditions. In addition, best way, long, travel times;
and hold were found to be the most efficient transition methods 2. The resulting impacts on networkwide and intersection-
for the base and 40% increased traffic conditions. For the 20% specific performance is quite different according to selected
increased traffic condition, best way performed the best. Based on EVP transition methods; and
these results, they concluded that the traffic volume condition 3. The dwell transition method often does not return the traffic
greatly influences the effectiveness of a given transition method. controller to coordination.
Cohen et al. 共2007兲 studied the performance of the dwell, max Although previous literature and research on EVP transition
dwell, subtract, and shortway methods on two different traffic methods have provided information and guidelines regarding the
networks using CORSIM Version 6.0. The test network, located performance and selection of individual methods, none of these
along Speedway Boulevard in Tucson, Ariz., contained ten inter- studies compared controller performance 共e.g., between 170 and
sections along a 7.1 km–long 共4.4-mi-long兲 corridor. The study 2070 controllers兲 nor considered the exit phase control available
results showed that delay and travel times under dwell and max in 2070 controllers. The research summarized in this paper at-
dwell transition methods spiked early in the transition process. In tempts to compensate for this lack of knowledge.
another hypothetical six-intersection network that operated near
oversaturated conditions, the results showed a similar spike in
delay under dwell and max dwell transition methods while sub- Case Study
tract and shortway allowed for a smoother return to coordination.
Cohen et al. concluded that, due to this “shockwave” in delay
Test Site and Data Collection
during the transition period caused by dwell and max dwell, those
methods are not appropriate for major-crossing thoroughfares. The test site, shown in Fig. 1, is a corridor along Lee Jackson
The above studies were based on transition methods embedded Memorial Highway 共U.S. Route 50兲 in Chantilly, Va., and in-

JOURNAL OF TRANSPORTATION ENGINEERING © ASCE / OCTOBER 2008 / 425

J. Transp. Eng., 2008, 134(10): 423-431


Downloaded from ascelibrary.org by AJOU UNIVERSITY on 03/23/23. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. Example of EVP operation

cludes four coordinated-actuated signals between Centerview brated to match the field traffic conditions after careful reviews of
Drive/Sullyfield Circle, and Chantilly Road. The dotted circles in the VISSIM manual and related references 共Planung Transport
Fig. 1 outline the four intersections under examination. They are Verkehr AG 2005; Park et al. 2006a兲. As the measure of effec-
referred to as Intersections 10–13 from the east of the network tiveness 共MOE兲 for calibration, the following two travel times
hereinafter as follows: were selected: 共1兲 westbound travel time from the eastern bound-
• Chantilly Road and Route 50 共Intersection 10兲; ary to the western boundary of the study area along Lee Jackson
• Metrotech Drive/Elmwood Street and Route 50 共Intersection Memorial Highway of 116.3 s; and 共2兲 eastbound travel time at
11兲; the same location of 102.2 s, which were both aggregated using a
• Centerville Road/Walney Road and Route 50 共Intersection 12兲; total of 16 and 15 individual travel time measurements, respec-
and tively, obtained during data collection with the GPS units. Finally,
• Centerview Drive/Sullyfield Circle and Route 50 共Intersection the VSSIM-based HILS was established with the resulting cali-
13兲. brated VISSIM network. The detailed results can be found in Best
Data collection efforts provided inputs for the VISSIM micro- 共2007兲.
scopic simulation program and performance measures for the
calibration of the VISSIM program, which was used in the evalu-
Traffic Signal Control and Preemption Settings
ation of EVP transition methods. While signal timings and link
geometry attributes were provided by the Virginia Department of As noted, the test site is currently being operated by McCain’s
Transportation 共VDOT兲, other data, including traffic and pedes- 170E traffic controllers. An EVP detector is installed at the north-
trian counts for network building and travel times for calibration, bound approach of Intersection 12 to serve the EVP requests from
were collected directly from the site on one weekday in 2006 the fire station located south of the intersection. The test site uses
between 3:00 p.m. and 4:00 p.m., which is referred to as the an optical preemption detector, which is mounted on the signal
off-peak traffic condition hereinafter. Traffic counts were mea- mast arm facing the northbound approach. Therefore, this study
sured using the JAMAR handheld manual traffic data collectors confines EVP capabilities to the northbound approach of Intersec-
and video recording using three cameras. For travel time mea- tion 12. The current EVP settings in the 170E controller were
surements, two vehicles equipped with global positioning system used:
共GPS兲 receivers traveled Lee Jackson Memorial Highway in both 1. Preemption function: emergency vehicle preemption;
directions collecting vehicle position data, which were then used 2. Preempt minimums:
to calibrate the VISSIM model. a. Pedestrian walk: 7 s;
b. Pedestrian flash don’t walk: 40 s; and
c. Minimum green: 10 s;
VISSIM Network and Calibration
3. Maximum preempt time: 255 s;
The VISSIM microscopic simulation program 共Version 4.10-14兲 4. EVP clearance time: 10 s;
was used in this study. Since the use of a well-calibrated micro- 5. EVP clearance Phases 4 共northbound through兲 and 7 共north-
scopic simulation model is a key to achieving reliable results for bound left turn兲; and
any simulation-based evaluations, the study network was cali- 6. Transition:
brated using a previously proposed microscopic simulation model a. Method: Shortway; and
calibration and validation procedure by Park and Qi 共2005兲. A b. Transition cycles: three transition cycles.
total of 11 calibration parameters, including car following behav- Based on the above settings, Fig. 2 illustrates an example of the
ior, lane changing behavior, and speed distributions, were cali- EVP operation at Intersection 12. In this example, it was assumed

426 / JOURNAL OF TRANSPORTATION ENGINEERING © ASCE / OCTOBER 2008

J. Transp. Eng., 2008, 134(10): 423-431


that an EVP call lasted for 20 s and was initiated at the local
controller timer’s 30 s mark, which is the middle of Phases 4 and
8. In Fig. 2, the resultant signal indications by the real timer,
master timer, and local timer were observed using the QuicLoad
software program 共BI Tran Systems 2000兲, which provides com-
munication between a computer and a 170E traffic controller via
a serial connection, and is used for monitoring controller perfor-
mance and transmitting data.
Upon receiving the EVP call, the controller returned to coor-
dination after 363 s. When the controller received the EVP call at
the real timer equal to 0 s, Phase 8 was terminated as soon as its
minimum green time was satisfied to serve the EVP phases 共i.e.,
Phases 4 and 7兲. After the EVP call was removed at the real timer
equal to 20 s, the controller showed 10 s of green indications—
Downloaded from ascelibrary.org by AJOU UNIVERSITY on 03/23/23. Copyright ASCE. For personal use only; all rights reserved.

the clearance time—to the EVP phases. The controller then


started the transition operation, in which the controller set the
master timer and the local timer equal to 0 s. When both reached
the background cycle length 共200 s兲, the controller reset the mas-
ter timer to its internal value as if an EVP call had not occurred
共i.e., at a real timer equal to 232 s, the master timer reached the Fig. 3. Physical and logical connections of HILS applied to test site
original cycle length of 200 s and then reset to 134 s兲. Mean-
while, the local timer continued to its adjusted cycle length ac-
cording to the selected transition method and number of transition
cycles. In the first cycle during transition, the controller increased
the cycle length up to 236 s, with the force-off points of all section 共i.e., 18 s兲 and the time to serve the preemption phases
phases adjusted in proportion to the increase in the total cycle after the EVP call was first made. If the latter was greater than the
length and Phase 3 skipped in the first cycle. In the second cycle, actual free-flow travel time due to the need to serve current pre-
the controller decreased the cycle length to 91 s, with the force- emption minimums 共e.g., minimum green and pedestrian walk
off points for all phases adjusted proportionally, and after which times兲, it was assumed that the emergency vehicle would not wait
the controller finally returned to coordination. at the intersection for a green indication, so the EVP call was
removed 5 s after the free-flow travel time of the emergency ve-
hicle elapsed to account for the vehicle weaving around traffic
Construction of VISSIM-Based HILS
and through the red light. Otherwise, the EVP call lasted for 18 s.
The VISSIM-based HILS consisted of the VISSIM microscopic Several studies found that the prevailing traffic condition dur-
simulation software, McCain/NITTA’s CID-II, and McCain’s ing EVP operations is one of primary factors that influences
170E controllers or Econolite’s 2070 controllers 共Planung Trans- which transition method is most effective 共Obenberger and Col-
port Verkehr AG 2005; BI Tran Systems 2000; McCain/NIATT lura 2001; 2007; Shelby et al. 2006兲. Therefore, this study inves-
2002兲. The parameters included controller settings 共e.g., mini- tigated the EVP transition methods available in the 170E and
mum greens, recall modes, vehicles extensions, pedestrian times兲, 2070 traffic controllers in different traffic volume conditions—
coordination settings 共e.g., cycle length, offsets, offset reference midday, off-peak, and p.m. peak. In addition to the differences in
point, green splits兲, and EVP settings 共i.e., preempt minimums traffic volumes among these conditions, the cycle lengths were
and preemption phases兲. Fig. 3 shows the physical and logical 150 s during the midday period and 200 s during the off-peak and
connections for the VISSIM-based HILS with the 170E traffic p.m. peak periods.
controllers.
170E Traffic Controller
EVP Transition Methods under Examination For the 170E traffic controller, the shortway method with two
different transition cycle settings 共i.e., two and three transition
During the preliminary analysis 共Park et al. 2006b兲, it was found cycles兲 were tested using the current phase sequence—lead-lead
that transition operations vary according to the location in the left turns with overlap on the cross streets at Intersection 12.
cycle length when an EVP call occurs. To account for this impact, Currently, Phase 3 共southbound left turn兲 and Phase 7 共north-
this study selected 20 locations in the local cycle length for EVP bound left turn兲 are lead phases with overlap on the cross streets,
calls. For the first simulation run, the EVP call was placed at 10 s as shown in Fig. 4-1. In an attempt to improve the transition
after the termination of the green indications for Phases 2 and 6 operation, this study examined an alternative phase sequence—
共i.e., the yield point兲 following the 6 min simulation warmup lead-lag left turns with overlap on the cross streets, as shown in
time. The EVP call for the subsequent simulation runs occurred at Fig. 4-2. In the lead-lag left turns with overlap sequence, Phase 3
10 s increments, allowing for the 20 total runs to span the entire was served after Phase 4 in an attempt to alternate the sequence of
200 s cycle length in separate VISSIM-based HILS runs. In ad- phases during the transition period. In this sequence, the green
dition, to consider the variability in the VISSIM microscopic traf- intervals used in the lead-lag left turns case were identical to
fic simulation model linked to HILS, three randomly seeded those of the lead-lead left turns case.
replications were made for each location. Thus, a total of 60 HILS The dwell transition method, which is available in the 170E
runs were made for each scenario. controller, was not considered for this study because transition
The length of an EVP call was determined by the assumed operations in a previous study did not return the controller to
free-flow travel time of the emergency vehicle through the inter- coordination in 38% of cases when using dwell 共Park et al.

JOURNAL OF TRANSPORTATION ENGINEERING © ASCE / OCTOBER 2008 / 427

J. Transp. Eng., 2008, 134(10): 423-431


Downloaded from ascelibrary.org by AJOU UNIVERSITY on 03/23/23. Copyright ASCE. For personal use only; all rights reserved.

Fig. 4. Phase sequence on cross streets

2006b兲. The cases presented in Table 2 were selected for the Implementation of Emergency Vehicle Preemption
evaluation of the EVP transition methods under three different
traffic conditions. To initiate and terminate EVP calls, this study used a software
program built in C# that has the following capabilities:
2070 ATC Traffic Controller 1. Initiates a VISSIM simulation run;
Unlike the 170E controller, the 2070 controller is able to employ 2. Transmits an EVP call to the subject controller after the
exit phases in an attempt to minimize the impacts of EVP by simulation warmup time according to user-specified con-
adjusting the phase sequence after the EVP operation. A portion straints;
of the study focused on determining the most effective exit 3. Terminates the EVP call;
phases. Phases 2 and 6 共eastbound and westbound through move- 4. Terminates the VISSIM simulation run; and
ments兲, Phases 4 and 8 共northbound and southbound through 5. Saves VISSIM-generated output files.
movements兲, Phases 3 and 8 共southbound left turn and through In the VISSIM-based HILS, 6 min of warmup time and
movement兲, and Phases 1 and 5 共westbound and eastbound left 20 min of simulation time were used. The actual simulation run
turns兲 were considered as exit phases. In this experiment, smooth time of 20 min was sufficient to consider the entire period of EVP
was used as the transition method because previous studies found operations, clearance time, and transition over the maximum
it to be the most effective 共Nelson and Bullock 2000; Obenberger number of cycles. The resultant performance measures were ex-
and Collura 2001; Shelby et al. 2006兲. tracted only during the simulation time.
Once the best exit phases were identified, smooth, add only,
and dwell with three different maximum dwell times 共100, 200,
255 s兲 were evaluated under the three different traffic volume Evaluation of EVP Transition Methods
conditions. Table 3 summarizes five EVP transition methods
evaluated in this study. Evaluation of EVP Transition Methods with
170E Traffic Controller
The EVP transition methods available in the 170E controller,
Table 2. EVP Transition Methods Using 170E Traffic Controller shown in Table 2, were evaluated using the VISSIM-based HILS.
Case number Cross street phase sequence Transition cycles Table 4 presents the performance measures from the evaluation of
a
these methods. These include the eastbound and westbound travel
170-1 Lead-lead left turns with overlap 3 times, networkwide average vehicle delay, and average number of
170-2 Lead-lead left turns with overlap 2 stops.
170-3 Lead-lag left turns with overlap 3 Under off-peak traffic conditions, Case 170-3 共shortway with
170-4 Lead-lag left turns with overlap 2 three transition cycles, lead-lag left turns兲 was found to be the
a
Current controller setting at the test site most effective EVP transition method by producing the best net-
workwide performance measures, as shown in Table 4.
Table 3. EVP Transition Methods Using 2070 ATC Traffic Controller
Under midday traffic conditions, there were no statistically
Case number Transition method Dwell time significant differences among the four cases according to Tukey’s
2070-1 Smooth — method’s of analysis of the two main measures—networkwide
2070-2 Add only —
average delay and travel time along the main corridor. However,
Case 170-4 共shortway, two transition cycles, lead-lag left turns兲
2070-3 Dwell 100
showed the least amount of networkwide average delay.
2070-4 Dwell 200
Under p.m. peak traffic conditions, Case 170-3 共shortway,
2070-5 Dwell 255
three transition cycles, lead-lag left turns兲 showed the least

428 / JOURNAL OF TRANSPORTATION ENGINEERING © ASCE / OCTOBER 2008

J. Transp. Eng., 2008, 134(10): 423-431


Table 4. Travel Times and Networkwide Performance Measures Using 170E Controller
Average travel time Networkwide performance
共s兲 measures
Average Avg number of
Traffic Case delay stops
condition number Eastbound Westbound 共s/veh兲 共stops/veh兲
a c
Off-peak 170-1 132 共19.2兲 149 共13.2兲 76 共6.5兲 2.1 共0.3兲
170-2 135 共19.4兲 151 共13.0兲 77 共6.5兲 2.1 共0.3兲
170-3 125 共18.1兲 142 共17.7兲 74 共7.4兲 2.1 共0.4兲
170-4 128 共16.6兲 144 共12.9兲 75 共6.1兲 2.1 共0.3兲
Midday 170-1 143 共15.0兲 186 共27.4兲 95 共9.6兲 3.6 共0.7兲
170-2 142 共10.8兲 177 共22.3兲 91 共6.8兲 3.3 共0.5兲
Downloaded from ascelibrary.org by AJOU UNIVERSITY on 03/23/23. Copyright ASCE. For personal use only; all rights reserved.

170-3 144 共14.8兲 175 共25.2兲 92 共9.6兲 3.4 共0.7兲


170-4 142 共11.2兲 173 共16.6兲 91 共6.5兲 3.3 共0.5兲
p.m. peak 170-1 185 共22.5兲 179 共19.0兲 105 共11.0兲 4.7 共0.8兲
170-2 193 共27.5兲 185 共18.3兲 109 共11.8兲 4.9 共0.8兲
170-3 185 共24.1兲 177 共17.2兲 105 共11.7兲 4.7 共0.9兲
170-4 187 共26.3兲 183 共21.1兲 106 共12.3兲 4.8 共0.9兲
a
Case 170-1 is the current EVP setting at the test site.
b
The calculation of the average time to coordination was conducted using only one run of each case 共20 replications with different EVP locations兲. The
time to coordination is the elapsed time from the start of the transition operation to coordination.
c
Numbers in parenthesis 共 兲 = standard deviation.

amount of networkwide average vehicle delay, but there was no Evaluation of EVP Transition Methods with 2070 ATC
significant difference between Cases 170-1 共shortway, three tran- Traffic Controller
sition cycles, lead-lead left turns兲 and 170-4.
Comparing the cases with different transition cycle settings, an Selection of Exit Phases
optimum number of transition cycles for the controller to return to The EVP transition methods for the 2070 controller, shown in
Table 3, were evaluated using the VISSIM-based HILS. As men-
coordination could not be identified. Comparing the different
tioned, this section begins with identifying the best exit phases
phase sequences on the cross street, as shown in Fig. 4, the tran-
following EVP. Four exit phase sets as well as a case without
sition operations using leadlag left turns with overlap performed specified exit phases were tested under off-peak traffic conditions,
better than those using lead-lead left turns with overlap. The im- as shown in Table 5.
pact of changing the phase sequence can be shown in a compari- As expected, it was found that all cases with exit phases out-
son of the average and maximum red intervals of Phase 3. For performed the no exit phase case. By exiting to nonpreemption
example, under the off-peak traffic conditions, Cases 170-3 共347 s phases, the traffic signal can immediately serve approaches that
of maximum red interval for Phase 3兲 and 170-4 共356 s of maxi- likely contain queued vehicles 共e.g., Phases 2 and 6—the coordi-
mum red interval for Phase 3兲 showed less average delay and nated phases兲. In the no exit phase case, however, the traffic sig-
shorter red intervals by phase at Intersection 12 than Cases 170-1 nal continues serving the EVP phases and proceeds with the
共470 s of maximum red interval for Phase 3兲 and 170-2 共i.e.,
482 s of maximum red interval for Phase 3兲. As illustrated in Fig.
2, this discrepancy occurred because the lead-lead left turn phase Table 5. Travel Times and Networkwide Performance Measures by
sequence skipped Phase 3 共southbound left turn兲 in the first cycle Selected Exit Phases
during transition. In the lead-lag left turn phase sequencing, how- Average
ever, the controller was able to give a green indication to Phase 3 travel time Networkwide performance
in the first cycle of transition by changing the phase sequence on 共s兲 measures
the cross streets. This result indicates that the effect of changing Average Average number
the phase sequence 共i.e., exit phase selection兲 is significant. Be- delay of stops
cause the 170E controller does not have the functionality to Exit phases Eastbound Westbound 共s/veh兲 共stops/veh兲
specify exit phases after preemption operations, however, its per- 4&8 117 共13.7兲b 125 共13.3兲 68 共4.9兲 1.8 共0.2兲
formance after EVP is limited. In addition, it was identified that a 3&8 117 共12.1兲 125 共14.4兲 68 共4.3兲 1.8 共0.2兲
kind of exit phase control by changing the phase sequence as 1&5 115 共4.7兲 123 共4.9兲 67 共2.5兲 1.8 共0.1兲
shown in Case 170-3 is able to give additional benefit beyond the 2&6 111 共8.6兲 117 共5.0兲 66 共2.4兲 1.7 共0.1兲
EVP parameter control, including the selection of different tran- No exit phasea 117 共13.0兲 129 共13.8兲 74 共4.6兲 1.8 共0.2兲
sition methods and transition cycle numbers. However, the im- a
The travel times, delay, and number of stops of the No exit phase case
pacts depend on traffic conditions during the preemption were extracted from only 40 replications.
b
operations as evidenced by the other two traffic conditions. Numbers in parenthesis 共 兲 = standard deviation.

JOURNAL OF TRANSPORTATION ENGINEERING © ASCE / OCTOBER 2008 / 429

J. Transp. Eng., 2008, 134(10): 423-431


Table 6. Travel Times and Networkwide Performance Measures Using 2070 ATC Controller
Average travel time Networkwide performance
共s兲 measures
Average
Case delay Average number of stops
Traffic condition number Eastbound Westbound 共s/veh兲 共stops/veh兲
Off-peak 2070-1 111 共6.3兲a 117 共5.0兲 66 共2.4兲 1.7 共0.1兲
2070-2 135 共11.6兲 139 共11.2兲 73 共3.7兲 2.0 共0.2兲
2070-3 103 共4.2兲 111 共4.4兲 64 共4.2兲 1.6 共0.1兲
2070-4 110 共4.6兲 118 共6.1兲 70 共2.7兲 1.7 共0.1兲
2070-5 110 共5.2兲 119 共4.6兲 70 共3.5兲 1.8 共0.1兲
Midday 2070-1 156 共7.2兲 144 共8.2兲 78 共3.5兲 3.1 共0.3兲
Downloaded from ascelibrary.org by AJOU UNIVERSITY on 03/23/23. Copyright ASCE. For personal use only; all rights reserved.

2070-2 162 共5.2兲 161 共8.3兲 81 共2.0兲 3.1 共0.3兲


2070-3 154 共6.3兲 152 共11.3兲 83 共4.6兲 3.2 共0.3兲
2070-4 157 共14.1兲 156 共16.8兲 87 共8.4兲 3.4 共0.6兲
2070-5 164 共14.5兲 157 共21.1兲 89 共10.7兲 3.6 共0.7兲
p.m. peak 2070-1 164 共15.6兲 158 共13.9兲 99 共8.5兲 4.2 共0.7兲
2070-2 194 共16.4兲 193 共17.8兲 112 共5.6兲 5.0 共0.5兲
2070-3 166 共9.1兲 165 共14.7兲 111 共6.8兲 4.1 共0.5兲
2070-4 163 共8.2兲 158 共9.3兲 103 共5.9兲 4.4 共0.5兲
2070-5 160 共9.5兲 159 共11.7兲 100 共4.8兲 4.1 共0.4兲
a
Numbers in parenthesis 共 兲 = standard deviation.

regular phase sequence while transitioning back to coordination. Under p.m. peak traffic conditions, no one transition method
Based on this finding, the importance of applying the exit phase stood out statistically in any of the three main MOEs, but Case
control to the EVP transition methods could be emphasized. 2070-1 共smooth兲 showed the least networkwide average delay. It
Among the four exit phase cases, the multiple t-test using was thus found that the best EVP transition method may be dif-
Tukey’s method indicated that no single method was outstanding ferent according to the traffic volume conditions and the time-of-
in terms of networkwide average delay 共Mann 2004兲. Exit Phases day 共TOD兲 plans.
2 and 6, however, showed significant improvements in eastbound The results of the experiments also show the advantages of the
and westbound travel time over the other exit phases. As a result, 2070 controller over the 170E controller in transitioning after
Phases 2 and 6 were chosen as the exit phases for the EVP ex- EVP. For example, under off-peak traffic conditions, the compari-
periments using the 2070 ATC controllers. son between Case 170-3—the best 170E EVP transition method
from the previous section—and Case 2070-3 indicates the advan-
Evaluation of Transition Methods tages of using the 2070 controller, which is capable of specifying
Transition methods available in the 2070 ATC controller were exit phases after EVP. Case 2070-3’s eastbound and westbound
then studied using the selected exit phases. Five cases were es- travel times were lower by 22 s 共17.6%兲 and 31 s 共21.8%兲, re-
tablished for the three different transition methods—smooth, add spectively, than Case 170-3. The networkwide average vehicle
only, dwell with 100 s dwell time, dwell with 200 s dwell time, delay was also lower by 10 s 共13.5%兲.
and dwell with 255 s dwell time, as shown in Table 3. Three
HILS runs, each with 20 different locations in the local cycle
length for the EVP call, were made for each case. Table 6 shows
the results. Conclusions and Recommendations
Under off-peak traffic conditions, the experimental results
show that Case 2070-3 共dwell with the maximum dwell time of This study compared various EVP transition methods using
100 s兲, performed best in networkwide average delay, eastbound VISSIM-based HILS at an urban corridor along Lee Jackson Me-
travel time, and westbound travel time by approximately 2, 6, and morial Highway 共U.S. Route 50兲 in Chantilly, Va.
7 s, respectively, over the second-most effective scenario, Case In a comparison of the transition methods available in the
2070-1 共smooth兲. Case 2070-3’s improvements were realized 170E controller, Case 170-3 共shortway with three cycles and lead-
most clearly at the noncoordinated movements—the northbound, lag left turns with overlap on the cross streets兲 showed the best
southbound, eastbound left, and westbound left approach delays networkwide and intersection-specific performance measures
were much lower than the other cases while eastbound and west- under off-peak traffic conditions. No one transition method stood
bound approach delays were very similar among all cases except out under midday and p.m. peak traffic conditions, however.
Case 2070-2. In addition, by exiting to the coordinated phases, the The evaluation of the transition methods available in the 2070
travel times of the major street were also reduced, as shown in controller identified that Case 2070-3 共dwell with 100 s of maxi-
Table 6. These findings support those of Shelby et al. 共2006兲. mum dwell time and exit Phases 2 and 6兲 yielded the best net-
Under midday traffic conditions, Case 2070-1 共smooth兲 out- workwide and intersection-specific performance measures under
performed the other cases by significant margins, improving by off-peak traffic conditions. However, Case 2070-1 共smooth兲 per-
approximately 3 and 8 s of average network delay and westbound formed the best under midday traffic conditions. Unde p.m. peak
travel time, respectively, over the second-best transition method. traffic conditions, no single transition method stood out statisti-

430 / JOURNAL OF TRANSPORTATION ENGINEERING © ASCE / OCTOBER 2008

J. Transp. Eng., 2008, 134(10): 423-431


cally in any of the three main MOEs, while Case 2070-1 共smooth兲 References
yielded the least networkwide average delay.
Based on the evaluation results of this study, the following Best, M. 共2007兲. “Evaluation of traffic controller algorithms after emer-
conclusions were made: gency vehicle preemption.” MS thesis, Univ. of Virginia, Charlottes-
1. The use of exit phases during the transition to coordination is ville, Va.
BI Tran Systems, Inc. 共1994兲. Model 170E controller traffic signal pro-
one of the key factors in the EVP transition operation. By
gram 233 VA, Sacramento, Calif.
changing the phase sequence in the 170E controller and BI Tran Systems, Inc. 共2000兲. QuicLoad user’s manual, Sacramento,
using the exit phase control in the 2070 ATC controller, dis- Calif.
ruptions to normal traffic signal operations after EVP can be Cohen, D., Head, L., and Shelby, S. G. 共2007兲. “Performance analysis of
significantly mitigated; and coordinated traffic signals during transition.” Proc., 86th Annual
2. The minimax class of transition methods 共e.g., smooth, short- Meeting of the Transportation Research Board, TRB, National Re-
way兲 generally perform best, which is similar to previous search Council, Washington, D.C.
Econolite Control Products, Inc. 共2004兲. ASC/2070 software program-
findings in the literature. In addition, the performance of
Downloaded from ascelibrary.org by AJOU UNIVERSITY on 03/23/23. Copyright ASCE. For personal use only; all rights reserved.

ming manual, Econolite Control Products, Inc., Anaheim, Calif.


transition methods may be dependent on the TOD plan and Mann, P. S. 共2004兲. Introductory statistics, Minitab 14 manual, 5th Ed.,
traffic conditions when EVP occurs. Therefore, the EVP tran- Wiley, New York.
sition method should be selected by carefully considering the McCain/NIATT. 共2002兲. Controller interface device (CID II): User’s
TOD plans in use and prevalent traffic conditions. manual, version 2.0, McCain Traffic Supply, Vista, Calif.
Based on the findings from this study, several recommenda- Nelson, E. J., and Bullock, D. 共2000兲. “Impact of emergency vehicle
tions were made: preemption on signalized corridor operation: An evaluation.” Trans-
portation Research Record. 1727, Transportation Research Board,
1. The cycle lengths used in the evaluations were 150 s under Washington D.C., 1–11.
off-peak traffic conditions and 200 s under peak traffic con- Obenberger, J., and Collura, J. 共2001兲. “Transition strategies to exit pre-
ditions. Given that short cycle lengths 共i.e., less than 100 s兲 emption control: State-of-the-practice assessment.” Transportation
could impact EVP performance and operations differently, Research Record. 1748, Transportation Research Board, Washington,
future research should investigate this variable; D.C., 72–79.
2. Since different exit phases worked better for certain locations Obenberger, J., and Collura, J. 共2007兲. “Using software-in-the-loop simu-
lation methodology to evaluate traffic signal transition strategies em-
of EVP calls within the local cycle, it is recommended that
ployed to exit preemption control.” Proc., 86th Annual Meeting of the
controller vendors add a feature allowing dynamic selection Transportation Research Board, TRB, National Research Council,
of exit phases. This would help to improve transition opera- Washington, D.C.
tions after EVP; Park, B., and Qi, H. 共2005兲. “Development and evaluation of simulation
3. There is a need to examine the impact of transition opera- model calibration procedure.” Transportation Research Record. 1934,
tions when an EVP call occurs on the mainline. Preliminary Transportation Research Board, Washington, D.C., 208–217.
experimental results on mainline EVP calls showed that the Park, B., Won, J., and Yun, I. 共2006兲. “Application of microscopic simu-
lation model calibration and validation procedure: A case study of
impacts on the mainline were more significant than those on coordinated actuated signal system.” Transportation Research Record.
a side street. Also, the role of exit phase control in the use of 1978, Transportation Research Board, Washington, D.C., 113–122.
the 2070 controller was more substantial 共Park et al. 2007兲. Park, B., Yun, I., and Best, M. 共2006兲. “Evaluation of emergency vehicle
There is another need for further research effort; and preemption strategies on a coordinated actuated signal system using
4. The use of HILS in the comparison of the EVP strategies has hardware-in-the-loop simulation.” Draft Rep., Center for Transporta-
shown great potential in the evaluation of EVP operations. tion Studies, Univ. of Virginia, Blacksburg, Va.
Park, B., Yun, I., and Best, M. 共2007兲. “Evaluation of pre-emption and
However, the huge computational time of HILS is a signifi- transition strategies for Northern Virginia smart traffic signal systems
cant obstacle to overcome. As an alternative, the use of SILS 共NVSTSS兲.” Final Draft Rep., Virginia Transportation Research
may be considered as the technology becomes more avail- Council, Va.
able, as suggested by Obenberger and Collura 共2001兲. Planung Transport Verkehr AG. 共2005兲. VISSIM 4.10 user’s manual,
Berlin.
Shelby, S. G., Bullock, D., and Gettman, D. 共2006兲. “Transition methods
in traffic signal control.” Transportation Research Record. 1978,
Transportation Research Board, Washington, D.C., 130–140.
Acknowledgments
Yun, I., Best, M., and Park, B. 共2007兲. “Evaluation of emergency vehicle
preemption strategies on a coordinated actuated signal system using
The writers thank Northern Virginia VDOT personnel for their hardware-in-the-loop simulation.” Proc., 86th Annual Meeting of the
support during data collection and for providing information on Transportation Research Board, TRB, National Research Council,
existing EVP strategies. Washington, D.C.

JOURNAL OF TRANSPORTATION ENGINEERING © ASCE / OCTOBER 2008 / 431

J. Transp. Eng., 2008, 134(10): 423-431

You might also like