Professional Documents
Culture Documents
1 s2.0 S2046043017300138 Main
1 s2.0 S2046043017300138 Main
Active signal priority control method for bus rapid transit based
on Vehicle Infrastructure Integration q
Li Zhou, Yizhe Wang ⇑, Yangdong Liu
The Key Laboratory of Road and Traffic Engineering, Ministry of Education, Tongji University, Shanghai 201804, China
a r t i c l e i n f o a b s t r a c t
Article history: The implementation of signal priority control to reduce delays of BRT vehicles at signalized
Received 2 February 2017 intersections is of practical and theoretical significance. In this paper, we propose an active
Received in revised form 1 June 2017 signal priority control method for BRT vehicles that run on median-road exclusive BRT
Accepted 4 June 2017
lanes at single intersections based on Vehicle Infrastructure Integration system. This
Available online 16 June 2017
method aims at maximizing average passenger benefit of BRT and other road users, and
provides 8 signal priority control scenarios respectively for 8 BRT arrival modes that are
Keywords:
based on estimating BRT vehicle travel time and locating arriving time window in a cycle.
Intelligent transportation system
Active signal priority control
The delay, energy efficiency and passengers’ comfort of BRT vehicles, and community vehi-
Bus rapid transit cles’ efficiency are also being considered. Finally, a model simulation was conducted by
Vehicle Infrastructure Integration VISSIM modeling in a representative signalized intersection with BRT in Jinan City,
VISSIM China. The results indicate that the proposed method reduces average passenger delay
by 13.43–25.27% and improves travel speed of BRT vehicles by 7.10–7.55% comparing to
existing signal control scenarios. The proposed method is highly promising and can be
applied to improve efficiency and safety of BRT at signalized intersections.
Ó 2017 Tongji University and Tongji University Press. Publishing Services by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
licenses/by-nc-nd/4.0/).
Introduction
Bus Rapid Transit (BRT) is a highly effective traffic mode that can relieve urban traffic pressure with its high quality, high
efficiency, low energy consumption and low cost when comparing with other traffic modes. However, unlike the subway
systems, the operation of BRT is subject to signal timing at the intersections that BRT vehicles travel though. The design
of traffic signal phases, circle length and delay is directly affecting the operation quality of BRT.
In the situation of continuously increased car travel ratio and deteriorated urban congestion and exhaust pollution, it is of
great practical significance to improve the service quality of BRT with advanced technologies. Vehicle Infrastructure Integra-
tion (VII) exactly brings new advances to improve the operational efficiency of BRT. VII is aiming at obtaining and integrating
comprehensive transportation information and implementing the coordinated control of people, vehicles and roads in order
to improve road capacity and relief traffic congestion, in the utilization of advanced technologies such as large-scale parallel
computing and sensor networks. BRT vehicles can be given signal priority when passing through an intersection with VII’s
implementation of certain signal priority control algorithms such as green extension, red truncation, and special phase inser-
tion, so that the operational efficiency of BRT system can be significantly improved.
Peer review under responsibility of Tongji University and Tongji University Press.
⇑ Corresponding author.
E-mail address: 16wangyizhe@tongji.edu.cn (Y. Wang).
http://dx.doi.org/10.1016/j.ijtst.2017.06.001
2046-0430/Ó 2017 Tongji University and Tongji University Press. Publishing Services by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
100 L. Zhou et al. / International Journal of Transportation Science and Technology 6 (2017) 99–109
Despite the great prospects of VII, the application of this technique still has immature parts especially the parts of signal
priority control methods and algorithms. Some researches only found numerical methods for signal priority control and sys-
tematic results are still in a shortage. Therefore, this study proposes a real-time signal priority control algorithm for single
intersections which can achieve the precise priority control of BRT vehicles that run on median-road exclusive BRT lanes,
according to the information interaction between the vehicle and the signal controller. The proposed algorithm is based
on VII to obtain vehicles information of location and speed precisely so that the travel time of BRT vehicles can be precisely
predicted.
Literature review
The increasing proportion of bus rapid transit (BRT) delay at signalized intersections in total travel time has made the
field of improving of BRT’s level of service (LOS) a research hotspot with practical and theoretical significance. Chen et al.
(2008) focused on the design and evaluation of BRT signal priority simulation under mixed traffic flow conditions.
Zlatkovic et al. (2012) analyzed different transit signal priorities (TSPs) for a future BRT corridor in West Valley City, Utah
to find the optimal TSP strategy from no-TSP, TSP, TSP with phase rotation, and custom TSP strategies. He finally found that
Custom TSP would provide major benefits for BRT in travel times, delays, and stops. Alomari et al. (2016) evaluated the per-
formance of various BRT with and without TSP along International Drive (I-Drive) in Orlando, Florida and simulation results
showed that TSP and BRT scenarios were effective in reducing travel times (up to 26%) and delays (up to 64%), as well as
increasing the speed (up to 47%), compared to the base scenario.
BRT Signal priority contains two critical parts – travel time prediction and active priority control. In the field of travel
time prediction, Balke et al. (2000) proposed a prediction model to predict the travel time of a BRT vehicle to arrive at a bus
stop and stop line at a signalized intersection, and whether to implement TSP can be determined based on travel time
prediction. The results showed a significant decrease of travel time with TSP. Ma et al. (2007) studied the relationships
between bus frequency, signal cycle and number of arrival-time-point, and obtained optimal number of arrival-time-
point by analyzing their effects on the deviation of bus average delay and headway. Kim and Rilett (2005) developed an
improved TSP algorithm that explicitly considers the prediction interval to reduce the negative impacts of nearside bus
stops. They used weighted-least-squares regression models to estimate bus stop dwell time and the associated prediction
interval, and tested the proposed TSP algorithm on a VISSIM model. It was found that the proposed TSP algorithm
improved bus operations without statistically significant impacts on signal operations. Kumar et al. (2015) proposed a
model-based algorithm motivated by the Kalman filter identifying the optimum inputs to predict bus travel time using
GPS data. A case study was conducted on two selected bus routes in the city of Chennai, India and the results obtained
from the algorithm are promising and showed the prediction accuracy to be within ±5 min for a prediction window of
30 min during 92% of the instances.
In the field of active priority control, Yang et al. (2001) proposed a bus signal priority control method within fixed
cycle. The proposed method determines green time of each phase by minimizing total passenger delay. Ling and
Shalaby (2004) proposed an adaptive traffic signal priority strategy based on Reinforcement Learning (RL). The RL con-
ducts priority by allowing transit vehicles to recover to the scheduled headway. Janos and Furth (2002) developed a
bus priority control strategy that responds quickly to and recovers quickly from priority interruptions, which is applicable
to the case of high bus arrival frequency. Nichols and Bullock (2004) analyzed the upper bound of benefits brought by
active bus priority. To reduce the impact of TSP on community vehicles, Balke et al. (2000) and Wu et al. (2013) used
real-time GPS data of bus to predict arrival time to stop line at intersections and stops, and accordingly determine
whether to switch priority phase for buses. Wang et al. (2014) established a cooperative bus priority system in connected
vehicle environment. The system was deployed and validated at an intersection with two adjacent bus stops in the city of
Taicang, China, and the results indicated that the proposed system could reduce travel time and decrease the number of
stops. Zeng et al. (2014) proposed a stochastic mixed-integer nonlinear program (SMINP) model to implement real-time
TSP control.
It can be seen from literature review that many researches focused on models or algorithms of bus priority that are devel-
oped using numerical or simulating methods, which have not yet form systematic results. For example, the lack of imple-
menting technology or application scenarios prevents priority algorithms from being directly applied to real operation
situation of BRT system. Besides, the researches are not adequate with regard to coordination between public transit and
community vehicles. BRT priority can make a significantly negative impact on community vehicles in some cases, even result
in crashes that are evitable without priority.
Real-time location and timetable deviation of BRT can be directly obtained from GPS system equipped on each BRT vehi-
cle, so that priority control can be implemented according to schedule adherence of BRT. The proposed priority control
method is based on the following considerations.
L. Zhou et al. / International Journal of Transportation Science and Technology 6 (2017) 99–109 101
The process of active BRT priority control method at single intersections which is shown in Fig. 1. contains the following
steps:
Step 1. The Road Side Unit (RSU) of Vehicle Infrastructure Integration located upstream of the approach at a signalized
intersection detects whether a BRT is arriving, if yes then priority control system predicts medium value of arrival
time at stop line tTTA and boarder spacing r, thus to calculate arrival time window and then go to STEP 2; if no then
remains present signal scheme.
Step 2. Locate lower bound, median value and upper bound of arrival time window respectively in a cycle, then go to STEP 3.
Step 3. Determine arrival modes of BRT and go to STEP 4. There are 8 BRT arrival modes classified from various arrival cases
and each mode has correspondent priority control strategy which is extension of green, early end of red, or insertion
of new phase.
Step 4. Determine operation condition of BRT using GPS data. Go to STEP 5 if BRT is behind schedule, or not behind schedule
but meets the condition of priority control. Otherwise, remain present signal scheme and end the algorithm.
Step 5. Calculate PI value within and without the implementation of priority control respectively. If the difference of PI
value is in an acceptable range then implement priority control according to the arrival mode determined in STEP
3, else remain present signal scheme and end the algorithm.
Furth and Muller (2000) made an in-depth research of conditional priority control and its method has been applied in
Eindhoven, Netherlands. They found that the delay of community vehicles in the cases of non-priority and unconditional pri-
ority is relatively larger than that in conditional priority. Therefore, PI values that represents overall traffic efficiency needs to
be calculated respectively for non-priority and priority control cases before implementing priority. And priority should be
applied only when the difference of PI values is within a certain acceptable range, else original signal scheme should remain.
PI values are calculated in units of per capita delay. And the difference of PI values before and after implementing priority DPI
can be calculated in Eq. (1)
DPI ¼ a DPIv þ b DPIt ð1Þ
where
DPI—difference of overall traffic efficiency at single intersection, seconds (hereinafter referred to as ‘‘s”);
a—weight for difference of community vehicles’ efficiency;
DPIv—difference of community vehicles’ efficiency, s;
DPIt—difference of BRT’s efficiency, s;
b—weight for difference of BRT’s efficiency.
102 L. Zhou et al. / International Journal of Transportation Science and Technology 6 (2017) 99–109
In Eq. (1) a and b can be set according to different management demands. And priority is implemented only when DPI is
in an acceptable range shown in Eq. (2).
e 6 DPI 6 w ð2Þ
where
e—the acceptable lower bound of overall traffic efficiency at single intersection, s;
w—the acceptable upper bound of overall traffic efficiency at single intersection, s.
Calculation of PI value
Red truncation
where
Green extension
t TTA þ r g t
DPIt ¼ Mi ð4Þ
2r
where
" #
Mi m X
1
DPIt ¼ ðg i þ Ii Þ þ g mmin ðtTTA rÞ M i ½r t ðt TTA rÞ ð5Þ
2r i¼1
where
Community vehicles
X
nþ1
0
DPIv ¼ ðNi dv i Ni di Þ ð6Þ
i¼1
where
Arrival time window at stop line of a BRT vehicle can be predicted immediately when it is detected by RSU located
upstream of the approach at a signalized intersection. Firstly, the distance from the location of a BRT vehicle to stop line
should be large enough such that travel time of BRT from the location detected by RSU for the first time to stop line remains
larger than system response time. In this paper, the distance between BRT vehicle’s location detected by RSU for the first
time and stop line is assumed to be 150 m. Next, linear model is used to predicted arrival time window, in which predicted
arrival time and its standard deviation are respectively calculated as median value and deviation of the upper or lower bound
to median value. For BRT that has exclusive right of way, travel time prediction model is shown in Eq. (7).
tTTA ¼ t s þ t t ð7Þ
where
tTTA: median value of predicted BRT arrival time at stop line;
tt: BRT travel time from the location to stop line, s;
ts: BRT stop dwell time, s.
104 L. Zhou et al. / International Journal of Transportation Science and Technology 6 (2017) 99–109
The variance of travel time prediction model can be calculated using Eq. (8).
1
r2 ¼ s2d 1 þ þ s2a1 ðx1i x1 Þ2 þ þ s2an ðxni xn Þ2 ð8Þ
n
where
r2: variance of travel time prediction model;
Sd: sample standard deviation of dependent variable;
SA: sample standard deviation of independent variables;
n: the number of independent variables, set to be 2 in this paper;
xi: sample value of independent variables;
x: sample mean of independent variables.
If Eq. (8) is inappropriate to calculate the variance and standard deviation in some cases, standard deviation can be
set as a varying parameter (e.g. 3 s) in simulation experiments. Arrival time window of BRT can be obtained using tTTA
and r2 in Eq. (9).
The median, lower bound and upper bound of BRT arrival time window needs to be located in a cycle and in phases. We
use the following steps to conduct localization, which can be represented in Eq. (10).
(1) Add the predicted travel time (in units s) to present time;
(2) Subtract integral multiple of cycle length from (1)’s result. Then we get the location of arrival time window in a cycle;
(3) Determine the phases that locations of the lower bound, median value and the upper bound of arrival time window
are on.
Furthermore, as shown in Fig. 2, eight arrival modes can be classified according to the localization of arrival time in dif-
ferent phases. Each arrival mode has a correspondent priority strategy since the lower bound, median value and the upper
bound are on different phases in a cycle.
Eight priority control strategies corresponding to all eight arrival modes mentioned in the previous section are illustrated
in Fig. 3.
The correspondent priority strategy is red truncation, in which the reduction length (shown in Eq. (11)) should be larger
than minimum green time of the previous phase of BRT priority phase.
Fig. 3. Eight priority control strategies each for an BRT arrival mode.
" #
X
n1
Dtred ¼ min r; r ðg i þ Ii Þ g nmin ð11Þ
i¼1
where
Dtred : reduction length of red time in BRT priority phase, s;
r: deviation of lower bound to the median value in a BRT arrival time window, s;
r: red length of BRT priority phase, s;
n: total number of phases except BRT priority phase;
g i : green length of each phases except BRT priority phase in original signal scheme, s;
Ii : green interval of each phases except BRT priority phase, s;
g n min : minimum green time in previous phase of BRT priority phase, s.
The correspondent priority strategy is green extension, in which the extension length (shown in Eq. (12)) should meet
BRT’s travel demands and extended green time should be less than maximum green time of the phase.
where
Dtgre : extension length of green time in BRT priority phase, s;
g t max : maximum green time in BRT priority phase, s;
g t : green time of BRT priority phase in original signal timing, s.
106 L. Zhou et al. / International Journal of Transportation Science and Technology 6 (2017) 99–109
The correspondent priority strategies for arrival mode 4 and 5 is the same green extension, in which the extension length
(shown in Eq. (13)) should meet BRT’s travel demands and extended green time should be less than maximum green time of
the phase. In Eq. (13) 2r is for meeting the travel demands of BRT vehicles.
where
tins —the insert time for green light;
m—the phase which for the lower bound of arrive time;
g m min —the minimum green light time in phase m;
yi —the ratio of community vehicles in phase i.
The priority control method for mode 7 is the reduction of red light time, the reduction time for red light is:
" #
X
n1
Dtred ¼ min r ðt TTA rÞ; r ðg i þ Ii Þ g nmin ð15Þ
i¼1
Model simulation
A model simulation was conducted based on the data from No. 1 BRT line at the intersection of Beidayuan Road and Wuy-
ingshan Road in Jinan City, China to test the control effects of proposed algorithm. As shown in Fig. 8, we implemented the
proposed algorithm through secondary development in C# and the algorithm was then tested in simulation experiments on
VISSIM5.2 after integrated though COM interface. This simulation contains 4 groups of experiments. Traffic flow of commu-
nity vehicles on the entrance lanes on arterial road (containing exclusive BRT lanes) was set as 600, 800, 1000 and 1200 pcu/
h respectively for each group of the simulation, while that flow on the entrance lanes on sub-arterial road was set constantly
as 600 pcu/h for all the groups of simulations (see Fig. 4).
(1) The average delay reduced 7.1 s, 6.3 s, 6.3 s and 6.7 s respectively when the traffic flow is 600, 800, 1000 and 1200 pcu/
h. Average passenger delay decreased by 13.43–25.27% with the increase of PI value. The result is shown in Fig. 5;
(2) As shown in Fig. 6, the speed of BRT vehicles has increased 7.12%, 7.10%, 7.55% and 7.35% when the traffic flow is
respectively the 4 setting values above;
108 L. Zhou et al. / International Journal of Transportation Science and Technology 6 (2017) 99–109
(3) We conducted 226 times of simulation experiments when the traffic flow was set as 1000 pcu/h. Results indicates that
all arrival modes appeared during our simulation and the appearance numbers for the modes are shown in Fig. 7. We
find that the appearance of arrival mode 6 (green light phase insertion) only accounts for 3.98%, which means that
mode 6 has considerable negative effect on the operation of traffic at a signal control intersection, such as the increase
of community vehicles’ delay and safety problem. The better solutions are green extension (arrival mode 3, 4, and 5)
with total proportion 38.94%, and red truncation (arrival mode 1, 7, and 8), with total proportion 34.96%. Therefore,
arrival modes of green extension and red truncation are preferred as optimal signal priority strategies.
To sum up, the simulation results show several positive effects brought by the proposed signal priority control algorithm
for BRT, including the decrease of average personal delay and the increase of BRT vehicles’ operational speed, and indicating a
promising application prospects of the proposed algorithm.
Conclusion
Vehicle Infrastructure Integration system is the new trade for transport operation, especially for the entire urban trans-
port system. This paper focused on the signal priority control method for BRT at single intersections, and designed working
schedule and kernel algorithm for it, presented some real-time priority control ways according to the arrive time of BRT in
different signal cycle, which has theoretical and practical value synchronously. The results indicate that the proposed
method reduces average passenger delay by 13.43–25.27% and improves travel speed of BRT vehicles by 7.10–7.55% com-
paring to existing signal control scenarios. In the further, the priority control for multiple intersections and stop station will
be the further research field for this paper.
References
Alomari, A., Al-Deek, H., Sandt, A., Rogers, J.H., Hussain, O., 2016. Regional evaluation of bus rapid transit with and without transit signal priority. Transp.
Res. Rec. J. Transp. Res. Board 2554 (18).
Balke, K.N., Dudek, C.L., Thomas Urbanik, I.I., 2000. Development and evaluation of intelligent bus priority concept. Transp. Res. Rec. J. Transp. Res. Board
1727 (1727), 12–19.
Chen, X., Yu, L., Zhu, L., Yu, L., Guo, J., 2008. Microscopic simulation approach to effectiveness analysis of transit signal priority for bus rapid transit: a case
study in Beijing. Transp. Res. Rec. J. Transp. Res. Board 2072 (2072), 64–76.
Furth, P., Muller, T.H., 2000. Conditional bus priority at signalized intersections: better service with less traffic disruption. Transp. Res. Rec. J. Transp. Res.
Board 1731, 23–30.
Janos, M., Furth, P., 2002. Bus priority with highly interruptible traffic signal control: simulation of san juan’s avenida ponce de leon. Transp. Res. Rec. J.
Transp. Res. Board 1811, 157–165.
Kim, W., Rilett, L., 2005. Improved transit signal priority system for networks with nearside bus stops. Transp. Res. Rec. J. Transp. Res. Board 1925, 205–214.
Kumar, B.A., Mothukuri, S., Vanajakshi, L., Subramanian, S.C., 2015. Analytical approach to identify the optimum inputs for a bus travel time prediction
method. Transp. Res. Rec. J. Transp. Res. Board 2535, 25–34.
Ling, K., Shalaby, A., 2004. Automated transit headway control via adaptive signal priority. J. Adv. Transp. 38 (1), 45–67.
Ma, W., Yang, X., Yun, M., 2007. Transit signal priority strategies based on the consideration of bus frequency. J. Tongji Univ. 21 (5), 353–366.
Nichols, A., and Bullock, D., 2004. Planning procedures for estimating an upper bound on bus priority benefits. In: Applications of Advanced Technologies in
Transportation Engineering, Eighth International Conference, pp. 169–174.
Wang, Y., Ma, W., Yin, W., Yang, X., 2014. Implementation and testing of cooperative bus priority system in connected vehicle environment: case study in
Taicang City, China. Transp. Res. Rec. J. Transp. Res. Board 2424, 48–57.
Wu, W., Ma, W., Yang, X., 2013. Simultaneous Optimization of Signal Timings, Departure Time and Dynamic Speed for Bus Priority Operations. In:
Transportation Research Board 92nd Annual Meeting, (No. 13–3537).
Yang, X.G., Lin, Y., Hang, M.S., 2001. Study of solution for transit priority signal. China J. Highway Transp. 14 (S1), 101–104.
Zeng, X., Zhang, Y., Balke, K.N., Yin, K., 2014. A real-time transit signal priority control model considering stochastic bus arrival Time. IEEE Trans. Intell.
Transp. Syst. 15 (4), 1657–1666.
Zlatkovic, M., Stevanovic, A., Martin, P.T., Tasic, I., 2012. Evaluation of transit signal priority options for future bus rapid transit line in west valley city, utah.
Transp. Res. Rec. J. Transp. Res. Board 2311 (-1), 176–185.