Available online at www.sciencedirect.com

*%4 ScienceDirect

Procedía Social and Behavioral Sciences 16 (2011) 450-459

6th International Symposium on Highway Capacity and Quality of Service Stockholm, Sweden June 28 - July 1, 2011

Analysis of Queue Formation and Dissipation in Work Zones

Hani Ramezani a*, Rahim Benekohalb

aResearch Assistant, Department of Civil and Environmental Engineering, University of Illinois, Urbana Champaign,USA bProfessor, Department of Civil and Environmental Engineering, University of Illinois, Urbana Champaign,USA Champaign,USA

Abstract

Studies on queue length and delay estimation have considered work zone as a single bottleneck. This work is the first study that acknowledges the presence of two locations that can be considered as potential bottlenecks: The work space and the transition area. Mechanism of queue propagation and dissipation are investigated by applying the shockwave theory on several initial traffic conditions that may happen in work zones. The study will address conditions under which these two bottlenecks will be functioning independently and when they interact with each other. Queue propagation and dissipation are estimated for a work zone using field data and the results are compared with the queue lengths observed in the field. © 2011 Published by Elsevier Ltd.

Keywords: Work zone, Shock wave, Bottleneck, Queue;

1. Introduction

Knowing queue length and delay at highway bottlenecks is critical in traffic management and design. Results of the queuing analysis are used in deciding the hours of work zone operation (peak, off peak, daytime, night time), selecting detours, making temporary capacity improvements, or providing real-time information to motorists.

Some past studies focused on delay and queue length estimation in work zones. Jiang (1999) and Chitturi et al. (2008) developed models to estimate delay and users' cost in work zones where vehicles are stopped in the queue. Ramezani et al. (2010) proposed a methodology to estimate moving queue length and corresponding delay in work zones. Chitturi and Benekohal (2009) modeled effects of speed distribution and speed difference between cars and trucks on delay estimation. Furthermore, they developed speed-flow curves for uncongested condition using simulation data and proposed a step-by-step methodology to estimate delay.

The above-mentioned studies assumed that there was only one bottleneck location in a work zone that caused the queuing and congestion. The queuing and congestion sometimes started at the transition area (Jiang 1999), or at the work space (Chitturi et al. 2008, Chitturi and Benekohal 2009 , Ramezani et al. 2010). This is the first study that acknowledges, there are two locations in work zones, namely the work space and the transition area, which can be potentially operating as a bottleneck.

* Corresponding author. Tel.: +1-217-819-2670. E-mail address: hrameza2@illinois.edu.

1877-0428 © 2011 Published by Elsevier Ltd. doi:10.1016/j.sbspro.2011.04.466

This paper uses shockwave analysis to investigate queue formation and dissipation in work zones where there are one or more bottlenecks. In Section 2, the geometry of a typical work zone is explained. Section 3 introduces notations. Sections 4 and 5 explain the mechanism of queue formation and dissipation, respectively. In Section 6, queue formation and dissipation is estimated using shockwave theory for a real work zone and the results are compared with the queue lengths, observed in the field. Conclusions and discussions will be made in Section 7.

2. Geometric of a typical 2-1 work zone

This section describes a typical and its simplified version of a 2-to-1 work zone, where one of the two lanes is closed due to construction activities.

0 ^^ Work Space

Buffer space

Advanced Transition Termination

warning area area Activity area area

Figure 1: The typical sketch of 2-to-1 work zones

Figure 1 shows the typical sketch of a 2-to-1 work zone using the MUTCD definitions. The work zone consists of four areas: advance warning area, transition area, activity area and termination area. Activity area is further divided into buffer space and work space. Traffic moves from the left to the right in Figure 1.

As mentioned before, two locations can potentially be operating as a bottleneck: 1) the transition area where the capacity drops due to the lane closure 2) the work space where traffic slows down due to work activities and this speed reduction may cause a capacity drop. There may be many reasons for traffic to slow down near the work space. For instance, drivers may reduce speed to avoid any collision with workers, to respond to a flagger showing "Slow Down" paddle, or in response to other traffic control device.

For the purpose of problem formulation, it is assumed that lane drop happens abruptly at the beginning of the transition area. Figure 2 shows the simplified work zone sketch in which point C corresponds to the beginning of the transition area. Section 1 represents the work space, Section 2 shows the space between the beginning of the transition area and beginning of the work space, and Section 3 represents the two-lane section before the transition taper.

Section 3 Section 2 J J Section 1

Figure 2: Simplified sketch of a 2-to-1 work zone

3. Notation

The notations are defined below. Uia , Uib , and Uic= Points representing undersaturated traffic states on the fundamental diagram for Section i of the road.

C[= Point representing capacity on the fundamental diagram for Section i,

Qia, and Qib = Points representing congested traffic states on the fundamental diagram for Section i.

qXi = The flow rate corresponding to the traffic state Xi ux. = The operating speed corresponding to the traffic state Xi Sx.Yj = The shock wave created by the traffic states Xi and Yj.

4. Queue formation

Fundamental curves of the following sections of the road are needed for shock wave analysis:

1) Work space (between point A and B in Figure 2).

2) The one-lane section before work space (between point B and C in Figure 2)

3) Upstream of the transition area (before point C in Figure 2)

Since no work activity exists downstream of point A and also after a while, lane closure is terminated, it is assumed that the capacity of the roadway section downstream of A is higher than the capacity of the work space. Therefore, there is no potential bottleneck location after the work space and no need for the fundamental curve of this section.

Another assumption is that the capacity of the transition area is higher than the capacity of the work space. Otherwise, the departure volume of the transition area cannot exceed the capacity of the work space, and hence, there is practically just one bottleneck location, i.e. the transition area. In that case, the problem could be reduced to the simple case of a freeway section with a single bottleneck.

It is also assumed that capacity of the road before the transition area is higher than the capacity of the transition area. Thus, the following relationship exists among the capacities of the three sections: Cx < C2 < C3

Figure 3 shows the general form of the flow-density curves for Sections 1, 2, and 3.

Assume that a traffic wave U3 enters the two-lane section of the work zone. We keep track of the resulting shock waves and traffic state evolution over the work zone. The flow rate of the coming wave, qUs, will satisfy one of the following conditions:

Case 1, qUs < C-l < C2 < C3

Case 2, Cx < qUs < C2 < C3

Case 3, C-l < C2 < qUs < C3

Case 1 does not create any queuing condition since the flow rate of the incoming wave is less than the capacity of all the bottlenecks. Hence, queue propagation is studied only for Cases 2 and 3.

4.1 Case 2: Arriving volume less than the capacity of the transition area but greater than the work space capacity

In this case, the flow rate of the wave,U3b, is higher than C1 but less than both C2 and C3. Thus, it is expected to have queue in the work space. For the detailed analysis of the queue propagation, we need to know the initial traffic conditions over the work zone as wave U3b enters the work zone. Before introducing the initial conditions, the concept of "active bottleneck" is defined. An active bottleneck is a bottleneck whose discharge rate is not affected by downstream traffic condition (Daganzo 1997). The following initial conditions can cause queue propagation when a high volume wave such as U3b enters the work zone:

2.1) The work zone is in undersaturated condition,

2.2) The work space is the active bottleneck and the back of queue is in the one-lane section,

2.3) The work space is the active bottleneck and the back of queue is in the two-lane section

/ / \

/ \ \

! TW 4

v V-t \ \ X \ <йЕ

!/ ft / !/ / Mt // \\ \\ \\ \\ \\ \ \

Density

---Sectiorl: The work space

-SectiorZ: Betweenthe beginning rf the transition area and the beginning of the work space

—Section3: Two lane section, before the transition taper

Figure 3: Traffic states in case 2.1

Case 2.1: All the traffic states, created in Case 2.1 are shown in Figure 3. Temporal and special evolution of traffic states in the work zone are displayed in Figure 4. Assume there are undersaturated conditions of Ula, U2a, and U3a in the work zone. At some time, a traffic wave of U3b with high volume enters the work zone (See Figure 4a). The resulting shockwave, Su3bu3^ moves forward since the vector connecting U3a to U3b in Figure 3 has a positive slope.

When SU3bU3a reaches the transition area, the two-lane section will be completely under the state of U3b and a stationary shockwave, SUsbU2b, occurs at the beginning of transition area. As the state of traffic changes in the transition area, a forward shockwave of SUzbU2a is created (See Figure 4b). When SUzbU2a, reaches the work space, the process of backward queue building up begins since Cx < qUib = q^. Then, since the vector, connecting U2b to Q2a (Figure 3) has negative slope, the shockwave of SUzbQ2a, propagates backward as shown in Figure 4c. Moreover, the stationary shockwave of SQ2aCl occurs at the beginning of the work space. The wave Cx is created and the forward shockwave of SClUla moves along the work space until it arrives the end of the work space. Beyond the end of the work space, it is assumed that the capacity of the roadway increases and there is no queue to worry about in this study. On the other hand, when the backward shockwave SUzbQ2a reaches the lane drop location, a stationary shockwave, SQ3aQ2a, is created. Thereafter, the backward shockwave SUsbQ3a starts propagating through the two-lane section as displayed in Figure 4d.

C Partd B A

Figure 4: Traffic evolution in the work zone for case2.1

Cases 2.2 and 2.3: Cases 2.2 and 2.3 can be considered as subset of Case 2.1. In Case 2.2, it is assumed that the traffic wave with high volume encounters the back of the queue at the one-lane section. The traffic evolutions for Case 2.2 are those shown in Figure 4c -d. Similarly, Case 2.3, in which the back of queue is in two-lane section, is illustrated by Figure 4d. So there will be just one backward shockwave, SU3bQ3a, and no other moving shockwave will be generated if arriving volume stays at the same level.

4.2 Case3: Arriving volume is greater than the capacities of the transition area and the work space

In this case, it is assumed that the volume of the incoming wave is higher than the capacity the transition area. Like in Case 2, it is needed to consider the initial conditions for queuing analysis. The following initial conditions are considered:

3.1) the work zone is in undersaturated conditions,

3.2) the work space is the active bottleneck and the back of queue is in the one-lane section,

3.3) the work space is the active bottleneck and the back of queue is in the two-lane section,

3.4) the transition area is the active bottleneck and the back of the queue is in the two-lane section.

Case 3.1:Figure 5 displays traffic states in Case 3.1. Figure 6a illustrates the instant when the high volume wave of U3c enters the work zone with the undersaturated states of Ula, U2a, and U3a. The vector, connecting U3c to U3a (Figure 5) has positive slope, therefore the resulting shockwave of SU3cUsa moves forward until it reaches point C. At this time, one backward shockwave and one stationary shockwave are generated. Since the flow rate of State U3c is higher than the capacity of the transition area, C2, queue starts propagating backward (Figure 5) with a shock wave speed of SUscQ3b through the two-lane section. On the other hand, the stationary shock wave of SQ3bC2 and the wave of C2 are generated as shown in Figure 6b. After this instant, the arriving volume of the one-lane section (between B and C) is equal to the departure rate of the transition area. Besides,

forward (Figure 6c) over the one-lane section until it reaches the work space. As shown in Figure 5, the flow rate of State C2 is higher than the capacity of the work space, Cx, and another queue will propagate backward (Figure 6d).

Density

- S"':'ni l: The work space

-5ection2: Between the beginning if the transition area and the beginning of the work space

.......Section3: Two lane section, before the transition taper

Figure 5: Traffic states in case 3.1

Assuming that the time interval with high demand volume lasts long enough, backward shockwave SCzQ2a reaches the beginning of transition area, and then backward shockwave SQ3bQ3a starts moving through the two-lane section (Figure 6d).

One can conclude from Figure 5 that the speed of SQ3bQ3a is greater than that of SUscQ3b. Therefore, these two shock waves meet each other after a while and then SUscQ3a remains as the only moving shockwave through the work zone (Figure 6e).

■ State State U„ 1 State U„

c Part a B A

Figure 6: Traffic evolution in the work zone for case3.1

Case 3.2: In this case, the work space is the active bottleneck and the back of queue is in the one-lane section. Hence, the traffic condition at the upstream of the back of queue should be in undersaturated condition such that the its flow rate is less than the capacity of the transition area, C2, but greater than the work space capacity, Cx. Assume that traffic states at the upstream of the queue are U2b and U3b in the one-lane section and two-lane section, respectively. When the forward shockwave of SUsc Usb reaches the transition area, both the backward shockwave of SUscQ3b and the forward shockwave of SCzU2b start propagating. Figure 7b illustrates what happens shortly after these two shockwave were generated. Two separate queuing conditions exist in the work zone: one in the one-lane section and the other in the two-lane section. Forward shockwave SCzU2b and backward shockwave, SUzb Q2a move toward each other until they meet and the wave U2b, diminishes. From that time on, the evolution of traffic will be the same as those shown in Figure 6c-e as explained in Case 3.1.

Case 3.3: In this case, the active bottleneck is the work space and the back of queue is in the two-lane section as shown in Figure 7e and explained in Case 3.1

Case 3.4: In this case, the active bottleneck is the transition area. Hence, this case is the same as the condition displayed in Figure 7b and one can follow the rest of the evolution from Case 3.1.

Part a

Part b

Figure 7: Traffic evolution in the work zone for case 3.2

5. Queue dissipation

For the recovery condition, it is assumed that a low volume wave, for instance, U3a, enters the work zone

such that

qu3a < Cx < C2 < C3

It was shown in the queue propagation discussion that if the time horizon is long enough, the work space will be the only active bottleneck in both Case 2 and Case 3, and the back of queue reaches the two-lane section. Hence, this condition is considered as the initial condition.

Figure 8a shows the initial condition in the work zone when the wave of U3a meets the congested state of Q3a. The resulting forward shockwave, SUsaQ3a, moves forward until it reaches the transition area, point C. Then a stationary shockwave of SUsaU2aand a forward shockwave of SUzaQ2a are generated (Figure 8b). A similar process happens when SUzaQ2a reaches the work space. The stationary and forward shockwaves of SUzaUia and SUlaCal are generated, respectively (Figure 8c). After this time, the work zone is completely in undersaturated condition.

6. Application of the Model to Field Data

The mechanism of queue formation and dissipation was explained using shockwave theory in a typical 2-to-1 work zone. In this section queue length is computed using shockwave theory for a work zone on an Interstate

highway in the US. Then the results are compared with the queue lengths, observed in the field. First field data are explained and then details of computations are provided. Thereafter the findings are discussed.

Field data: Field data were collected from an interstate highway in Illinois (I-39NB). One of the two lanes was closed due to construction activities. A flagger with "Slow Down" paddle was present at the site and vehicles reduced their speeds in response to the flagger. Workers and construction equipments were working close to traffic and work intensity was high when queue was present. The extent of queue (queue length) was recorded at minute intervals by an observer. Beside queue length, headway, speed and type of each vehicle were recorded by videotaping. General information of the site is shown in Table 1.

Congestion lasted for about 25 minutes. The maximum queue length was 4680 ft and the queue did not propagate to the two-lane section. The estimated capacity and average speed during the congestion were 1064 pcphpl and 15.9 mph, respectively. Details of capacity calculation were explained by Benekohal et al. (2010).

Part b A

State || ...... 1 B f l^Ä

Part c

Figure 8: Queue dissipation in the work zone

Table 1: General Information of I-39NB

Capacity (pcphpl) Within-queue speed (mph) Percentage of trucks Speed limit (mph) Work intensity

1064 15.9 27 45 High

Computation: According to the field data, work zone was in undersaturated condition before the onset of congestion. Also no queue existed at the two-lane section. Hence the queue formation is similar to Case 2.1 for which arriving volume was lower than the capacity of the transition area but higher than the capacity of the work space.

Queue length at the end of each minute is estimated using shockwave theory and the results are compared with that of the field data. To compute shockwave speed, flow rate and density are needed for each traffic state. Flow rate, q, is available for each minute interval from field data and density, d, can be estimated using the speed, u, and flow rate, q, collected from the field:

d = Ru

Hence shockwave speed is computed as:

_ 1Q - 1u

Q=Represents the queuing condition,

U=Represents the undersaturated condition, upstream of queue in the one-lane section, SUQ=Speed (mph) of the shockwave created by the traffic states U and Q, qQand uQ = The flow rate (vph) and speed (mph) of the queuing condition, and qv and uu = The flow rate (vph) and speed (mph) of the undersaturated condition.

It is assumed that that qQ and uQ are fixed and equal to the capacity and the corresponding speed, respectively. Arriving volume during each minute, is also available and it was assumed that speed of vehicles upstream of queue is equal to free flow speed, 50 mph, which is five mph higher than speed limit. Since arriving volume is not fixed, the shockwave speed is computed for each minute and the corresponding queue length is estimated as:

QLj = T x SUQi + QLi_1

QLj and QLj_x are the estimated queue length at the end of each interval. QL0 is zero since there was no queue before the analysis interval. T=Interval length

SUQi=Shockwave speed during the interval i.

In order to have stable estimation, moving average queue length for interval i (MAQLj) are computed as:

QLj for the first and the last intervals

MA(^ = 1 QLj-! + QLj + QLi+1 .

--Otherwise

Findings: Figure 9 shows MAQLj's computed from field data versus those estimated using shockwave theory. The maximum overestimation and underestimation are about 400 ft and 1200 ft, respectively. Also, the difference between the peak of the estimated and observed queue lengths is about 195 ft (4.3% error) and those occur at minute 10. The average of the estimated queue lengths, 2200 ft, is not statistically different than that of the observed queue length, 2435 ft (P-value=0.73). Moreover, the error in the average of the estimated queue lengths is 9.6%. Based on the model, the total congestion period is about 23 minutes and this is close to the actual period which is 24 minutes. Hence the estimated queue propagation and dissipation reasonably describes the field data.

7. Conclusions and discussions

This is the first study that considers more than one bottleneck in work zones. The following bottlenecks may be active in a work zone: 1) the work space 2) the transition area 3) both. This paper analyzed the mechanism of queue formation and dissipation in work zones under various traffic conditions. .

Detailed analysis showed that when the volume exceeds capacities of the transition area and the work space, both locations will be the active bottlenecks. However, if the arriving volume maintains at the same or a higher level during a-long-enough interval, the back of the queue in the one-lane section of the work zone reaches the transition area, and the two bottlenecks will not be operating independently. After this time, the transition area will be deactivated as a bottleneck and the work space will be the only active bottleneck. Moreover, the queuedischarge rate may not be identical to within-queue flow rate. Nonetheless, in the long-run these two flow rates will be equal.

In addition, it was shown that when the arriving volume is less than the capacity of the transition area and more than the work space capacity, work space will be the only active bottleneck.

Queue propagation and dissipation was estimated using field data from a real work zone. The estimated queue lengths at the end of each minute interval were compared with those, observed in the field. The percentage of error in the estimated maximum queue length and the estimated average queue length were 4.3% and 9.6%

respectively. The average of the estimated queue lengths was not statistically different than that of the observed queue lengths. Therefore the estimated queue propagation and dissipation, reasonably describes field data. The actual mechanism of queue propagation and dissipation is complicated however this mechanism is simplified when we use macroscopic models to describe traffic states. In this case, stochastic elements in traffic such as the interaction between drivers or the presence of slow moving vehicles, are not considered any more.

No assumption was made about the general form of the fundamental curves, nevertheless; the analyses were made particularly for 2-to-1 work zones. One can use the results of this study to estimate delay, queue length, and extension of congestion as well as to detect the location of active bottlenecks in work zones.

— 5000

0 5 10 IE 20 25

Time Interval (min)

-MAQLfrom field MAQL using Shockwave theory

Figure 9: Moving Average Queue Length: Estimated Versus Observed

References

Benekohal, R. F.; Ramezani, H.; Avrenli, K.; 2010, Delay and User's Cost in Highway Work Zones, Research Report ICT-10-07. Retrieved from http://ict.illinois.edu/publications/report%20files/FHWA-ICT-10-075.pdf

Chitturi, M. V.;Benekohal, R. F.; Kaja-Mohideen, A.,2008, Methodology for Computing Delays and Users Costs in Work Zones, Transportation Research Record.

Chitturi, M.V., R.F. Benekohal, 2009.Work zone queue length and methodology. 89th TRB Annual Meeting CD-ROM.

Jiang, Y, 1999, A model for estimating excess user costs at highway work zones, Transportation Research Record, No. 1657.

Jiang, Y., 1999, "Traffic Capacity, Speed, and Queue-Discharge Rate of Indiana's Four-Lane Freeway Work Zones", Transportation Research Record 1657.

Daganzo, C., 1997. Fundamentals of Transportation and Traffic Operations. Elsevier, New York.

Ramezani, H.; Benekohal, R. F.; Avrenli, K.; 2010; Methodology to estimate moving queue length and delay in highway bottlenecks. TFT conference proceeding.