580 Summary
nwasal
DynamicSignalPriorityControlStrategytoMitigateOff-RampQueueSpillback.pdf
1
Transportation Research Record: Journal of the Transportation Research Board, No. 2438, Transportation Research Board of the National Academies, Washington, D.C., 2014, pp. 1–11. DOI: 10.3141/2438-01
Congestion downstream of a freeway off-ramp often produces a traffic queue at the mainline and thus reduces the freeway capacity at the interchange area. To prevent such queue spillback, a proposed real-time control strategy provides a priority signal control to the off-ramp traf- fic when potential queue spillback is detected. Two priority strategies— unconditional and conditional controls—that quickly discharge the queuing vehicles on the off-ramp can contend with different congestion patterns at the freeway and local arterial. Based on field data from a free- way interchange in Zhubei, Taiwan, extensive simulation experiments demonstrate the effectiveness of the proposed strategies in preventing off-ramp queue spillbacks. The results show that the overall network performance can be significantly improved with the proposed strategy, even though some roadway segments of the local arterial network may experience negative impacts.
Because the operational performance of a freeway segment and its neighboring local streets are often mutually dependent, a large body of research on concurrent control of the freeway and local arterial has been reported. Most such studies, however, focused on the on-ramp metering controls and their coordinated operations (1, 2).
The equally critical issue of off-ramp control, in contrast, has not yet received adequate attention. As described by Lovell, most drivers do not tend to segregate themselves by destination well in advance of an off-ramp, but make most of their lane-changing decisions at the last moment (3). The exit queue of an off-ramp might spread itself laterally upstream of an off-ramp, thereby restricting the efficiency of the mainline flows. Hence, congested conditions at downstream intersections can lead to a long traffic queue at the off-ramp, and the queue spillback may extend upstream and block freeway travel lanes (4–6). Considering a partial blockage of the right lane, Newell proposed a model to evaluate the delays on a freeway when queues from an exit ramp spill back to the freeway mainline (7 ). Cassidy et al. studied the exiting queue of an off-ramp with field data from videotapes and found that a bottleneck with a diminished capacity arose on a freeway segment whenever queues from a segment’s off-ramp spilled over and occupied its mandatory exit lane (8).
To mitigate freeway congestion caused by the excessive off-ramp queue, some studies proposed to use the lane assignment strategy
according to travelers’ destinations (9), and to detour the traffic flow to less congested ramps (10, 11). Hagen et al. studied the problem of queues at freeway off-ramps and developed a toolbox with a set of potential strategies to reduce the queues (12). Li et al. presented a mixed integer model for an integrated control between the off-ramp and arterial traffic flows, intending to minimize the queue spillback from an off-ramp to the freeway mainline (13). Similarly, Lim et al. proposed a signal control model to minimize the total delay for off- ramps and their connected arterials (14). Pei and Zhou developed a similar control model to optimize the green time and cycle length at a surface road on the basis of the off-ramp traffic conditions (15).
Obviously, the off-ramp flows will compete for the signal timings with local traffic. In addition, because of the fluctuation of freeway traffic and the stochastic arrival rate at the off-ramp, an integrated corridor control with pretimed strategies may not be sufficiently responsive. Using the real-time information, this study provides a control system to prevent the queue spillback at freeway off-ramps. Detectors are installed at the off-ramp and nearby local intersections, and a prediction model is employed to project the queue evolution in the next few signal cycles. Once the potential spillback is detected, a signal priority will be provided to the off-ramp flows, allowing the traffic to quickly pass the downstream intersections. In response to various traffic conditions at the freeway and local arterials, this study has developed two separate control strategies: unconditional and conditional priority control. Depending on the detected congestion patterns, one may select the proper priority strategy to minimize the off-ramp queue impact on the local traffic.
ReseaRch BackgRound
To formulate the interrelationships between the off-ramp queue and local traffic, this study selects a freeway segment in Zhubei, Taiwan, as the study site (see Figure 1a). During most evening peak hours, the freeway outflow rate to Guangming 6th Road is extremely high and often causes an oversaturated condition at the off-ramp. Because of the impact of downstream signals, the traffic queue on the off-ramp often builds up quickly and spills back to the freeway mainline. To record the traffic condition on the freeway segment through the peak hours, this study site has loop detectors upstream and downstream of the off-ramp entry to collect speeds and flow rates of approaching vehicles. A video camera has been installed near the bottleneck area for observation of the queue evolution during peak hours.
The traffic data were collected from 4:30 to 9:30 p.m. on April 25, 2013. Figure 1b shows the time-dependent speed profile on each lane upstream of the target off-ramp. From 5:00 to 8:30 p.m., one can observe significant speed reductions on all three freeway travel lanes.
Dynamic Signal Priority Control Strategy to Mitigate Off-Ramp Queue Spillback to Freeway Mainline Segment
Xianfeng Yang, Yang (Carl) Lu, and Gang-Len Chang
Department of Civil and Environmental Engineering, University of Maryland, College Park, 1173 Glenn L. Martin Hall, College Park, MD 20742. Corresponding author: X. Yang, [email protected].
2 Transportation Research Record 2438
The speed on Lane 3, nearest to the spillback lane, dropped to 20 km/h. However, after the vehicles pass the entry of the off-ramp auxiliary lane, their speeds on all three mainline lanes quickly recover to 90 km/h, as shown in Figure 1c. The presence of queue spillback during the congested period was also recorded by the video camera.
To improve operational efficiency, one can detour vehicles to other nearby off-ramps to reduce the off-ramp flows or adjust the down- stream signal timings to increase the discharging flows. The former, however, may not be applicable if nearby off-ramps are too far away. Hence, this study focuses on the latter strategy by dynamically adjusting downstream signals with a real-time control strategy. The research objective is to develop a proactive control strategy by pre- dicting queue evolution at the off-ramp and to offer a signal priority to prevent the off-ramp queue spillback. The priority control can be with or without preconditions, depending on the detected traffic states at the freeway and local intersections. To minimize the impacts on local traffic, the proposed off-ramp priority control includes a transition plan to smooth recovery to the original timing plan.
contRol FRamewoRk
The operational process for activating the priority control consists of the steps in the next sections.
online data collection
For the target freeway segment, the data collection is focused on recording the upstream traffic flow and speed and the outflow at the off-ramp. The upstream flow and speed can be used to identify the traffic condition on the freeway mainline, which serves as the basis to select the proper priority control strategy. The outflow at the off-ramp is used to estimate the queue evolution. Hence, one loop detector is installed at the freeway upstream and another detector is deployed at the entry of the off-ramp.
Detectors at local arterials are used to estimate the congestion level at nearby intersections. In this study, the congestion level is defined by the queue over link length ratio at each critical link. To estimate the queue length, detectors are installed upstream and downstream of each target local link within the impact area of the off-ramp flows.
Queue Prediction
To prevent queue spillback at the off-ramp, the system has devel- oped a prediction model to project the queue length evolutions. Given the potential queue spillback information, a control priority will be selected to discharge the off-ramp queues. The dynamic proactive control, with its queue prediction function, has two major advantages:
FIGURE 1 Study site: (a) National Highway 1, Zhubei, Taiwan; (b) speed obtained by upstream detectors; and (c) speed obtained by downstream detectors.
Bottleneck caused by queue spillback
Guangming 6th Road Loop
Detector Video
1 2 3
400 m 765 m
Time
4:30 p.m. 5:30 p.m. 6:30 p.m. 7:30 p.m. 8:30 p.m. 9:30 p.m.
20
0
40
60
80
100
120
S p e e d (
km /h
)
Lane 1 Lane 2 Lane 3
Time
4:30 p.m. 5:30 p.m. 6:30 p.m. 7:30 p.m. 8:30 p.m. 9:30 p.m.
20
0
40
60
80
100
120
S p e e d (
km /h
)
Lane 1 Lane 2 Lane 3
(a)
(b) (c)
Yang, Lu, and Chang 3
(a) it can activate the priority strategy before the formation of queue spillback and (b) it can minimize the impact on the arterial traffic by using the unconditional or conditional priority control.
Let Q(k) denote the distance between the end of queue and the stop line and q(k) be the point queue length, which equals the number of stopped vehicles in queue. Then, the time-dependent queue can be estimated with the following procedures:
( ) ( ) ( ) ( )+ = + −q k q k A k D k1 (1)
( ) ( ) ( )+ = + +Q k q k R k1 1 (2)
where A(k) is the number of arriving vehicles during time step k and R(k) is the distance between front of queue and stop line:
( ) ( ) ( ) ( )
( ) + =
+ + >
+ =
R k
R k D k q k
q k 1
if 1 0
0 if 1 0 (3)
D(k) is the number of departed vehicles during time step k and is defined as follows:
( ) = ∆
D k s t during green time
0 during red time (4)
p
where s is saturation flow rate and Δt is length of time interval. For an accurate prediction, it is essential to install a downstream
detector at the off-ramp; this detector can directly provide the departure information during the green time.
To predict the queue evolutions, this study employs a standard time series model using the historical detector data to execute the prediction function. Because the development of a time series model has been extensively reported in the literature, this paper will not discuss this module for the sake of condensing the model presentation.
congestion level evaluation
Considering the relationship between flow, density, and speed, the pro- posed control system uses the speed and flow rate obtained from the freeway upstream detector as the measurement of freeway congestion.
Because giving the signal priority to the off-ramp traffic flows may negatively affect other local traffic movements, the congestion levels at the affected links are defined by their ratios of queue over the link length. Given the upstream and downstream detector data in each target link, one can then estimate the queue length with Equations 1 through 3.
Priority decision
Given the congestion level and predefined thresholds at the freeway and local arterial, the priority decision module will be activated to determine the proper control strategy according to the following steps (see Figure 2).
Step 1. Freeway Congestion Level < Threshold
When the detected speed near the off-ramp is higher than the critical speed and the detected flow rate is below the threshold, the impact caused by queue spillback from the off-ramp will be limited. Hence, no priority control is required for this type of condition.
Step 2. Freeway Congestion Level > Threshold and Arterial Congestion Level < Threshold
If the estimated congestion level on the freeway exceeds the thresh- old while the arterial congestion level is below the threshold, the unconditional priority control strategy would be selected to dis- charge all queuing vehicles at the off-ramp. A detailed discussion of the unconditional priority strategy is given in the next section.
Step 3. Freeway Congestion Level > Threshold and Arterial Congestion Level > Threshold
If the congestion levels on both the freeway and local arterial exceed their respective thresholds, a conditional priority strategy, which can limit the potential negative impacts to the arterial flows, should be activated.
Figure 2 shows the flowchart for the entire control system.
dynamic PRioRity contRol stRategy
To effectively discharge queuing vehicles at the off-ramp, one essential step is to identify the primary path for the off-ramp flows and provide signal progression during the priority control period. Figure 3a illustrates the setting of signal progression for the off-ramp flows with the unconditional priority strategy. To formulate the green bandwidth for the off-ramp flow, the signal timings at all intersections within its impact boundaries should be adjusted by (a) extending the green time of the corresponding phase and (b) shifting the offsets to ensure the signal progression.
Because of the relatively large impacts on other traffic movements, the unconditional priority strategy should be used only when the congestion level at the local arterial is below the predefined threshold. Figure 3b shows the signal progression with the conditional priority strategy. Instead of discharging all queuing vehicles at the off-ramp to pass the downstream intersections, conditional priority strategy allows some residual queues at these intersections to minimize the impacts on other local traffic movements.
unconditional Priority control
The priority decision will be made when the potential queue spillback at the off-ramp is detected. The entire decision-making process for the unconditional priority control includes the following steps:
Step 1. Collect the pretimed signal plans at each local intersection within the affected boundaries; estimate the current queue length at the off-ramp.
Step 2. Compute the green bandwidth for progression of the off-ramp traffic movement. To discharge all queuing vehicles at the off-ramp, the system provides a sufficient green band for vehicles to go through all intersections on their primary paths. Given the current queue length at the off-ramp, the green bandwidth can be computed as follows:
=b q
s (5)1
where q1 is current queue length and s is saturation flow rate.
4 Transportation Research Record 2438
Step 3. Set the signal progression. For convenience of discus- sion, let the intersection connected to the off-ramp be denoted as Intersection 1. The key variables associated with progression design are shown in Figure 4.
Step 3.1. Compute the extended green time el and er. To pre- vent overflow at the off-ramp during signal transition, the extended green time at Intersection 1 will be determined by the proportion of red time at its left and right sides:
e R b g
R R l
l
l r
( ) =
− +
max , 0 (6)
1
e R b g
R R r
r
l r
( ) =
− +
max , 0 (7)
1
where
Rl = total red time at the left side of the green band, Rr = total red time at the right side of the green band, and g1 = original green time of the phase that provides right-of-
way to off-ramp flows.
Step 3.2. Compute the green time g–i for the off-ramp path at intersection i. To ensure traffic progression, the initial queue discharging time at the downstream links should be considered. Hence, the green time, g–i, is given by
g b q
s gi
j
j
i
i∑= +
=
max , (8) 2
where qj is the initial queue at the target link of intersection j. Step 3.3. Compute the time between the start of a cycle to the
left side of green band wi:
w R e tl l= − + − θ (9)2 1,2 2
w w t q
s ii i i i i i
i= θ − θ + + − >− − − 2 (10)1 1 1,
where ti−1,i is the travel time between intersection i − 1 and inter- section i and θi is the signal offset at intersection i.
Equations 9 and 10 show the relationships between signal tim- ings and travel time between adjacent intersections on the basis of
Level Evaluation
Freeway Data Collection
Volume Prediction Model
Off-Ramp Queue Prediction
Potential Spillback?
Freeway Congestion
Priority Decision
Arterial Data Collection
Arterial Congestion Level
Evaluation
Yes
No
Start i = 1
Cycle i Normal Signal Control
Data Collection
i = i + 1
Criterion Satisfied?
Yes
Unconditional Priority Control in Next Cycle
i = i + 1
No
i = i + 1
Conditional Priority Control in
Next Cycle
FIGURE 2 Flowchart of control system.
Yang, Lu, and Chang 5
these signals on the arterial being synchronized (i.e., they operate in a common cycle length). Step 4. Reassign the remaining green time to other phases. When
the green band of the priority path is extended, the green times of other phases need to be reduced to keep the cycle length unchanged. To balance the impact to each phase, the system will reduce the green
times with the same ratio if the minimum green time requirement is not violated. Hence, at each intersection, the objective function of the assignment plan is given as follows:
g g g
g g g
j j j
k k i
i i
k∑ ( )= − −
{ }∈δ ≠
(11)
,
where
δ = set of all phases, i = index of extended (target) phase, g–j = reassigned green time of phase j, and gj = original green time of phase j.
To reduce the negative impacts caused by the priority control strategy, the phase sequence should remain unchanged during the green time reassignment process.
Step 5. Design the signal transition. Given the computation results from previous steps, Figure 5 shows an example of signal transition between pretimed and real-time signal cycles. After one cycle of priority control is implemented in real time, the signal timings should
Queue Clearance Time
Queue Clearance Time
Extended Green Time
Off-Ramp Traffic
Time
Distance
Intersection 1
Intersection 2
Intersection 3
Queue Clearance Time
Queue Clearance Time
Off-Ramp Traffic
Residual Queue
Residual Queue
Residual Queue
Time
Distance
Intersection 1
Intersection 2
Intersection 3
(a)
(b)
FIGURE 3 Signal progression for off-ramp flow with (a) unconditional priority strategy and (b) conditional priority strategy.
el er
g1
gi
gi+1
bwi
wi+1
Intersection 1
Intersection i
Intersection i + 1
θi
θi+1
Rl Rr
FIGURE 4 Key variables associated with unconditional priority control strategy.
6 Transportation Research Record 2438
be returned to their original plan to mitigate the impacts on other local traffic movements.
conditional Priority control
As shown in Figure 3b, the conditional priority control strategy allows the presence of some residual queues at the downstream links. Similar to the unconditional priority control strategy, the decision-making process for such control includes the following steps:
Step 1. Collect the pretimed signal plans at each involved inter- section and estimate the current queue length at the off-ramp intersection and its downstream intersections;
Step 2. Determine the acceptable length of residual queue at each intersection; and
Step 3. Compute the green bandwidth for signal progression. As shown in Figure 6, compared with the unconditional priority
control strategy, the conditional strategy offers variable green bands between intersections. Hence, the green bandwidths between neighboring intersections are given by
= − ε
b q
s (12)1
1 1
b b
s i
i i= − ε− (13)1
where εi is the allowable residual queue at intersection i.
Step 4. Set the signal progression. Step 4.1. Compute the extended green time el and er. The
extended green time at Intersection 1 will be computed as follows:
e R b g
R R l
l
l r
( ) =
− +
max , 0 (14)
1 1
e R b g
R R r
r
l r
( ) =
− +
max , 0 (15)
1 1
Step 4.2. Compute the green time g–i for the off-ramp path at intersection i. In the conditional priority control strategy, the green time g–i at each intersection is given by
g b q
s gi i
j
j
i
i∑= +
=
max , (16) 2
Step 4.3. Compute the length of wi.
w R e tl l= − + − θ (17)2 1,2 2
= θ − θ + + − >− − − ; 2 (18)1 1 1,w w t q
s ii i i i i i
i
Note that θi could be a negative value. Step 5. Reassign the green time to other phases at each intersection.
This step is the same as the one for unconditional priority control. Step 6. Design the signal transition. The design of signal transition
is similar to the one for unconditional priority control.
case study
site description and data collection
To evaluate the proposed control system, one freeway segment in Zhubei, Taiwan, along with its nearby intersections was selected as the study site. As shown in Figure 7, during evening peak hours, heavy traffic volumes take the off-ramp via Node 4 to the Guangming 6th Road West. Because the current pretimed signal timings are
Off-Ramp Downstream Intersections
Time
Potential Queue
Spillover Queue
Prediction
Control System
Request
Local Signal Controller
Green Time Extension
Pretimed Cycle Real-Time Cycle Pretimed Cycle
Green Band
FIGURE 5 Signal transition between pretimed and real-time signal cycles.
FIGURE 6 Key variables associated with conditional priority control strategy.
e1 er
g1
gi
gi+1
b1 wi
wi+1
Intersection 1
Intersection i
Intersection i + 1
θi
θi+1
bi
bi+1
Yang, Lu, and Chang 7
designed to coordinate the through traffic on the arterial, the north- bound off-ramp flows often form a long queue and spill back to the freeway mainline.
To analyze the freeway capacity reduction caused by the queue spillback, the data collection team at National Chiao Tung University, sponsored by the Taiwan Department of Transportation, completed a field survey from 4:30 to 9:30 p.m. on April 24 and 25, 2013. The collected data include the following:
1. Freeway northbound flow rate, along with its turning ratios at the off-ramp (Node 4);
2. Traffic volume in each lane group at each intersection shown in Figure 8;
3. Maximum queue length per cycle at critical arterial links; using the node numbers in Figure 7, these critical links are Node 1 to Node 2, Node 2 to Node 3, Node 5 to Node 1, Node 8 to Node 3, Node 9 to Node 3, Node 10 to Node 3, Node 3 to Node 2, and Node 2 to Node 1;
4. Current signal timings, including cycle length, green splits, and offsets; and
5. The major path of the off-ramp movement: Node 4 to Node 10.
Vissim calibration
To evaluate the network performance before and after the online priority controls, this study produced a simulation network with VISSIM 5.20. Recognizing that a simulation system is useful only if it can faithfully reflect the behavior of its target driving populations, this study has performed the calibration by minimizing the difference between simulated and field collected queues as well as flow rates. The calibration results for VISSIM simulation are shown in Table 1 and in the list of adjusted parameters and values:
• Average standstill distance (urban), 3.22 ft; • Maximum deceleration (lane change), −14.99 ft/s2; • Accepted deceleration (lane change), −6.00 ft/s2; and • Maximum deceleration for cooperative braking, −14.99 ft/s2.
experimental Results
From the field observations, the major path for the off-ramp flows is from Node 4 to Node 10, as shown in Figure 7. Hence, the real-time
FIGURE 7 Geometric layout of study site (Nodes 1–10).
Guangming 6th Road East
Guangming 6th Road West
4
5
1
6
7
2
3
9
8
10
Xianzhen 2nd Road
Signal 1
Signal 2
Signal 3
17 s
19 s
34 s 52 s 46 s
19 s 30 s 14 s 63 s
48 s 19 s 42 s 17 s
4 s 2 s
Current Pretimed Signal Plans
8 Transportation Research Record 2438
priority control is designed to discharge the excessive traffic queue along this path.
To compare the network performance by control strategy, the experimental analysis includes the following three scenarios:
1. No control scenario. The signal timings are operated with the current fixed signal plans,
2. Unconditional priority scenario. Only the unconditional priority control strategy is implemented, and
3. Dynamic control scenario. The decision-making function will dynamically switch to the proper strategy according to the detected traffic conditions.
Following the control steps of each strategy, the real-time control frequency along with the assigned green band for off-ramp traffic are shown in Figure 8.
Evaluation of Freeway Mainline Performance
As shown in Figure 9, it is clear that under both real-time control strategies, the off-ramp queue did not spill back to the freeway mainline to cause negative impacts on the traffic flows; this result is evident through a comparison of the freeway travel times between the three scenarios. The experimental results with the simulation have shown that the proposed control strategies could yield satisfactory results and successfully discharge the off-ramp queue to prevent excessive traffic spillback.
Evaluation of Critical Traffic Paths
This subsection compares the travel time of several major local paths under the three control scenarios. The travel times from the off-ramp to the Zhubei area have been significantly improved with the priority control in the latter two scenarios, as evidenced by the results in Figure 10a. In addition, the travel times under the dynamic control strategy are slightly higher than those with unconditional control because the control produces a reduced green band when congestion is detected on local streets.
Figure 10b shows the travel time comparison for the path from Node 5 to Node 10. The travel time differences along this path under different control strategies are not significant. One possible reason is that the path has some shared links with the priority off-ramp path, and the clearance of queues at those shared links can also help reduce its travel time.
Paths that conflict with the off-ramp paths inevitably experienced an increase in total delays under the real-time strategies. As shown in Figure 10, c and d, the travel times for both southbound through and left-turn movements are significantly increased under the real-time control strategies. Compared with unconditional control, the travel times under the dynamic control are much shorter because it executes the conditional priority when long queues are detected at 6:30 p.m. on important local segments.
In conclusion, the real-time control strategies may result in dif- ferent levels of impact to critical traffic paths at the local arterial. To evaluate the network performance under different control strategies, Table 2 summarizes the analysis results, including average delay, average number of stops, and average speed.
The average delay is shown to decrease by 12.7% and 15.6%, respectively, under the unconditional and dynamic controls; the same trend could be found for the average number of stops.
Regarding the average network speed, the implementation of both priority control strategies can provide 7% and 8.7% improvements, respectively. Hence, one can confirm the effectiveness of both real- time control methods, and the one with dynamic control generally outperforms the other. Although some origin–destination paths under such controls may experience longer travel times, the traffic flows to the off-ramps and along the freeway mainline could gain significant improvement in average delays and speeds. Because the volume on the freeway is much larger than its conflicting counterparts, the total network could benefit from the proposed off-ramp strategies.
120
100
80
60
40G re
e n B
a n d (
s)
20
0 5:30 6:30
Time (p.m.)
7:30 8:30 9:30
Unconditional Control Dynamic Control
FIGURE 8 Provided green band for real-time control strategies.
TABLE 1 Percentage Difference Between Simulation and Field Volume Data
Percentage Difference, by Approach
Intersection Westbound Northbound Eastbound Southbound
1 1 0.6 2 na
2 0.9 na 2 0.2
3 2 3 0.6 1
Note: na = not applicable.
Yang, Lu, and Chang 9
FIGURE 9 Travel time passing bottleneck.
90
80
70
60
50
40
30
20
T im
e P
a ss
in g t h e B
o tt le
n e ck
( s)
10
0 4:30 5:30
Time (p.m.)
6:30 7:30 8:30 9:30
No Control Unconditional Control Dynamic Control
FIGURE 10 Travel time of paths: (a) Node 4 to Node 10 and (b) Node 5 to Node 10. (continued on next page)
900
800
700
600
500
400
Tr a ve
l T im
e (
s)
300
200
100
0 4:30 5:30
Time (p.m.)
6:30 7:30 8:30 9:30
Tr a ve
l T im
e (
s)
0
4:30
450
400
350
300
250
200
150
100
50
5:30
Time (p.m.)
6:30 7:30 8:30 9:30
No Control Unconditional Control Dynamic Control
(a)
No Control Unconditional Control Dynamic Control
(b)
10 Transportation Research Record 2438
To evaluate system performance under difference congestion levels, this study further tests three scenarios by decreasing or increasing flows on the entire network. Table 3 summarizes the results over the peak period (5:30 to 8:30 p.m.). The proposed system can improve the network performance in most cases. For the most congested case (flow + 10%), the high traffic flows have limited the operational effectiveness of the system because of the oversaturated condition.
conclusions
To mitigate freeway congestion caused by queue spillback at off- ramps, this study has proposed a real-time control model to discharge the traffic queue before it becomes excessive. The entire control sys-
TABLE 3 System Performance over Peak Hours, by Flow Pattern
Network Delay (s) Freeway Mainline Travel Time (s) Path Nodes 4–10 Travel Time (s)
Case Before After Improvement (%) Before After Improvement (%) Before After Improvement (%)
Current flow 105.4 92.9 −11.8 62.0 48.7 −21.5 545.1 337.4 −38.1 Flow (−5%) 78.6 73.4 −6.7 48.3 47.2 −2.3 327.1 259.7 −20.6 Flow (+5%) 168.5 97.0 −42.4 133.0 47.4 −64.4 690.6 307.4 −55.5 Flow (+10%) 197.6 194.8 −1.4 140.1 129.5 −7.6 723.8 598.6 −17.3
TABLE 2 Entire Network Performance
Performance Index No Control
Unconditional Only
Dynamic Control Strategy
Average delay (s) 89.065 77.77 (−12.7%) 75.209 (−15.6%) Average number of stops
2.391 1.711 (−28.4%) 1.621 (−32.2%)
Average speed (km/h) 36.116 38.633 (7.0%) 39.25 (8.7%)
FIGURE 10 (continued) Travel time of paths: (c) Node 8 to Node 9 and (d) Node 8 to Node 2.
No Control Unconditional Control Dynamic Control
Tr a ve
l T im
e (
s)
600
500
400
300
200
100
0 4:30 5:30
Time (p.m.)
6:30 7:30 8:30 9:30
No Control Unconditional Control Dynamic Control
Tr a ve
l T im
e (
s)
600
500
400
300
200
100
0
4:30 5:30
Time (p.m.)
6:30 7:30 8:30 9:30
(c)
(d)
Yang, Lu, and Chang 11
tem includes four core modules: data detection, queue length pre- diction, congestion level evaluation, and priority decision. To account for congestion levels at the freeway and local arterial, this study has presented both unconditional and conditional priority controls for contending with the off-ramp spillback issue. With one freeway seg- ment in Zhubei, Taiwan, as the study site, the proposed dynamic off- ramp control system has been tested with a well-calibrated VISSIM network. The base case has been compared with the proposed unconditional and dynamic controls, and the results of extensive simulation analysis have confirmed that both priority control strat- egies can successfully prevent the formation of queue spillback at the off-ramp. The dynamic control, with both unconditional and conditional priority strategies, can minimize the potential negative impacts to traffic flows at neighboring intersections.
This study is the first stage of the research. Other tasks include (a) identifying the potential operational issues associated with the proposed system, (b) designing a more comprehensive conditional control strategy to account for the operational efficiency of other key paths, and (c) implementing the system in the field and extensively evaluating its performance.
acknowledgment
The authors are grateful to the data collection group led by Ka Lo Wong at National Chiao Tung University for providing the field data used in the numerical example.
ReFeRences
1. Coordinated Operation of Ramp Metering and Adjacent Traffic Signal Control Systems. Turner–Fairbank Highway Research Center, FHWA, U.S. Department of Transportation, McLean, Va., June 1996.
2. Pooran, F. J., and H. C. Lieu. Evaluation of System Operating Strategies for Ramp Metering and Traffic Coordination. Proc., IVHS AMERICA 1994 Annual Meeting, Washington, D.C., 1994.
3. Lovell, D. J. Traffic Control on Metered Networks Without Route Choice. University of California, Berkeley, 1997.
4. Daganzo, C. F., M. J. Cassidy, and R. L. Bertini. Possible Explanations of Phase Transitions in Highway Traffic. Transportation Research Part A: Policy and Practice, Vol. 33, No. 5, 1999, pp. 365–379.
5. Muñoz, J. C., and C. F. Daganzo. The Bottleneck Mechanism of a Freeway Diverge. Transportation Research Part A: Policy and Practice, Vol. 36, No. 6, 2002, pp. 483–505.
6. Jia, B., R. Jiang, and Q. S. Wu. Traffic Behavior near an Off Ramp in the Cellular Automaton Traffic Model. Physical Review E, Vol. 69, 2004, p. 56105.
7. Newell, G. F. Delays Caused by a Queue at a Freeway Exit Ramp. Transportation Research Part B: Methodological, Vol. 33, No. 5, 1999, pp. 337–350.
8. Cassidy, M. J., S. B. Anani, and J. M. Haigwood. Study of Freeway Traffic near an Off-Ramp. Transportation Research Part A: Policy and Practice, Vol. 36, No. 6, 2002, pp. 563–572.
9. Daganzo, C. F., J. Laval, and J. C. Muñoz. Some Ideas for Freeway Con- gestion Mitigation with Advanced Technologies. Traffic Engineering and Control, Vol. 43, No. 10, 2002, pp. 397–403.
10. van den Berg, M., B. De Schutter, and H. Hellendoorn. Effects of On-Ramp and Off-Ramp Metering on Queue Forming in Urban Traffic Networks. Proc., 11th IFAC Symposium on Control in Transportation Systems, Vol. 11, Part 1, Aug. 2006, pp. 130–135.
11. Günther, G., J. E. Coeymans, J. C. Muñoz, and J. C. Herrera. Mitigating Freeway Off-Ramp Congestion: A Surface Streets Coordinated Approach. Transportation Research Part C: Emerging Technologies, Vol. 20, No. 1, 2012, pp. 112–125.
12. Hagen, L., P. S. Lin, and A. D. Fabregas. A Toolbox for Reducing Queues at Freeway Off-Ramps. Report BD544-10. Center for Urban Transporta- tion Research, University of South Florida, Tampa; Florida Department of Transportation, Tallahassee, 2006.
13. Li, Z., G. L. Chang, and S. Natarajan. An Integrated Off-Ramp Control Model for Freeway Traffic Management. Presented at 15th World Congress on Intelligent Transportation Systems, New York, 2009.
14. Lim, K., J. H. Kim, E. Shin, and D. G. Kim. A Signal Control Model Integrating Arterial Intersections and Freeway Off-Ramps. Journal of Civil Engineering, Vol. 15, No. 2, 2011, pp. 385–394.
15. Pei, Y., and K. Zhou. Off-Ramp Control near Surface Road. Research Journal of Applied Sciences, Engineering and Technology, Vol. 5, No. 8, 2013, pp. 2612–2615.
The Traffic Signal Systems Committee peer-reviewed this paper.
