Scholarly article on topic 'Real-time Queue Length Estimation of Signalized Intersections Based on RFID Data'

Real-time Queue Length Estimation of Signalized Intersections Based on RFID Data Academic research paper on "Civil engineering"

CC BY-NC-ND
0
0
Share paper
OECD Field of science
Keywords
{"traffic management" / "traffic signal" / "queue length" / "vehicle delay" / "RFID data"}

Abstract of research paper on Civil engineering, author of scientific article — Aoxiang Wu, Xiaoguang Yang

Abstract How to estimate queue length in real-time at signalized intersection is a long-standing problem. The traditional input–output approach for queue length estimation can only handle queues that are shorter than the distance between vehicle detector and intersection stop line, because cumulative vehicle count for arrival traffic is not available once the detector is occupied by the queue. Based on the analysis of traffic flow's shockwave profile on the approach of signalized intersections, the paper brings forward a model to detect the real-time queue length based on RFID detector data. The model solves the problem of measuring intersection queue length by exploiting the queue delay of individual vehicles instead of counting arrival traffic flow in the signal cycle. Variations of the estimation model are also presented under different traffic volumes, the relationship between queue length and the capacity of the approach is also under consideration. A field experiment using RFID Detector Data was conducted at an intersection in Nanjing. The results of field experiment demonstrate that the proposed models can estimate the real-time queue length with satisfactory accuracy.

Academic research paper on topic "Real-time Queue Length Estimation of Signalized Intersections Based on RFID Data"

Available online at www.sciencedirect.com -

ScienceDirect PrOC6d ¡0

Social and Behavioral Sciences

Procedia - Social and Behavioral Sciences 96 (2013) 1477 - 1484

13th COTA International Conference of Transportation Professionals (CICTP 2013)

Real-time Queue Length Estimation of Signalized Intersections

Based on RFID Data

Aoxiang Wu*, Xiaoguang Yang

Key Laboratory of Road and Traffic Engineering of the Ministry of Education and School of Transportation Engineering, Tongji University, _4800 Cao'an Road, JiadingDistrict, Shanghai, 201804, China_

Abstract

How to estimate queue length in real-time at signalized intersection is a long-standing problem. The traditional input-output approach for queue length estimation can only handle queues that are shorter than the distance between vehicle detector and intersection stop line, because cumulative vehicle count for arrival traffic is not available once the detector is occupied by the queue. Based on the analysis of traffic flow's Shockwave profile on the approach of signalized intersections, the paper brings forward a model to detect the real-time queue length based on RFID detector data. The model solves the problem of measuring intersection queue length by exploiting the queue delay of individual vehicles instead of counting arrival traffic flow in the signal cycle. Variations of the estimation model are also presented under different traffic volumes, the relationship between queue length and the capacity of the approach is also under consideration. A field experiment using RFID Detector Data was conducted at an intersection in Nanjing. The results of field experiment demonstrate that the proposed models can estimate the real-time queue length with satisfactory accuracy.

© 2013TheAuthors.PublishedbyElsevier Ltd.

Selectionandpeer-reviewunder responsibility ofChinese Overseas TransportationAssociation(COTA). Keywords:traffic management; traffic signal; queue length; vehicle delay; RFID data;

1 Introduction

Traffic management meets many critical challenges in most modern cities because the traffic congestion is continuing to grow in urban areas. It has long been recognized that the real-time queue length on signalized intersections is an important parameter for traffic management and control. During the past years, many researchers have dedicated themselves to the research of queue length estimation and a lot of models have been developed. The first one, which is based ra the analysis of cumulative traffic input-output to a signal link, was

* Corresponding author. Tel.: +86-18817308468; fax: 021-69589475. E-mail address: wuax.fish@gmail.com

1877-0428 © 2013 The Authors. Published by Elsevier Ltd.

Selection and peer-review under responsibility of Chinese Overseas Transportation Association (COTA). doi:10.1016/j.sbspro.2013.08.168

proposed by Webster (1958) and later improved by a number of researchers (Newell, 1965; Gazis, 1974; Sharma et al., 2007; Vigos et al., 2008). However, cumulative input-output techniques can only be used for the estimation of queue length when the rear of the queue does not exceed the vehicle detector, it cannot handle long queues exceeding beyond the detector, so applications of the technique are limited. Another important queue length estimation model based on the behavior of traffic shockwaves has been developed by Lighthill and Whitham (1955), foe theory was fast demonstrated for uninterrupted flow, and this model was later expanded by Stephanopolos and Michalopoulos (1979 and 1981) to signalized intersections. This model estimates queue lengths by tracing the trajectory of shockwaves based on the continuum traffic flow theory, it can successfully describes the complex queuing process in both temporal and spatial dimensions. Based on this theory, Henry X. Liu and Xinkai Wu (2009) proposed a method to estimate real-time queue length for congested signalized intersections using loop detector data, Xinkai Wu and Henry X. Liu (2010) proposed the identification of oversaturated intersections using high-resolution traffic signal data. Based on the theory of shockwaves, the process of vehicles queuing and dispersing at intersections and variety of traffic flow between intersections are studied by Wang Dian-hai and Jing Chun-guang (2002), they studied platoon-interval in artery of the city and affection of signal coordinate with considering the direction and speed of traffic wave spread, relevant mathematics models are founded in their paper. Fan Hong-zhe (2007) created a real-time queuing model and Grenberg model which is suitable for describe high-density traffic flow is used to modify it.

Almost all the researchers were use fixed detectors like loops, radar sensors, or digital cameras to detect the traffic information, the successful wide scale deployment of advanced traveler information systems (ATIS) and advanced traffic management systems (ATMS) depends on the ability to obtain and subsequently disseminate information that accurately reflects network traffic conditions, however, all of these approaches cannot identify the detailed information of a particular car. With the development of intelligent transportation system, RFID is more and more popular in traffic information collection. RFID applications have been studied in recent years as well, for example, Pico Holdings, a South African Company, developed and tested a passive RFID reader to monitor vehicles at a read distance of 17 feet traveling at speeds of 160 mph (RFID, 2004). When a car with a passive tag passes an RFID reader at a speed not exceeding 250 km/h, the car can be read accurately. W. Wen (2010) presents an intelligent traffic management expert system with RFID technology, the system provides both practically important traffic data collection and control information and can trace criminal or illegal vehicles such as stolen cars or vehicles that evade tickets, tolls or vehicle taxes. These studies are focus on using RFID technology to detect the link travel time and traffic status, but have no research on how to detect the traffic queue of signalized intersections.

In this paper, we develop an algorithm for the estimation of real-time queue length based on RFID data, the mechanism and implementation framework have been analyzed in accordance with shockwave profile of traffic flow in intersection. Several implementation suggestions have also been discussed. The filed test results prove the effectiveness of this model, it shows that the relative error of this model is under 15% in most of the time.

The remainder of this paper is organized as follows. In Section 2, the framework of the RFID system is introduced. In Section 3, the model to estimate queue length is described, and the test results are presented in Section 4. Finally conclusions are given in Section 5.

2 Framework of the RFID system

This research is based on the RFID system in Nanjing, Jiangsu province, China. The framework of the RFID system comprises a passive RFID tag, an RFID reader, a local area network (LAN) and a high-speed server with a database system (see Fig. 1). The passive RFID tag is pasted on the wind shield of a car, it doesn't have power, its power comes from the reader's RFID transmission. The reader detects data and sends it back to the high-speed server via a LAN, then storing it in the database, the server then calculates the average travel time and speed on

the link, to ensure the safety of information, we use wire LAN to transmission information instead of wireless LAN.

Fig. 1 Framework of the RFID system

In order to build this system, RFID tag had been pasted on over 30% of the cars in Nanjing, and there are more than 300 RFID detector points in the test area, each exit path of the intersection in the test area has a RFID detector (see Fig. 2).

Fig. 2 The location of RFID detector

The data sends back by the reader include tag ID, detector ID, detection time, car plate number and so on, to protected the security and privacy of private information, we only use the first four kinds of information (see Tab. 1).

Tab.1 RFID data

Tag ID Detector ID Time Car plate

317448BCC069222457***** 2774 2012/8/1 08:12:45 AB4***

3 The basic concept and shockwave profile analysis at signalized intersection

The analysis following is based on the identification of delay and the shockwaves presented in a cycle caused by traffic state changes. As indicated in Fig.3, a car is detected by the upstream RFID detector at time t0 (show as point A in Fig.3), the car is expected to arrive at the downstream detector at tj (show as point Q in Fig.3) without signal control and queue, but in fact, the car will stop at point B1 because of the queue and arrive at the downstream detector at t2 (show as point C2 in Fig.3) at last, so the total delay of the vehicle between the two detectors can be estimated by the following equation:

l + d (1)

Delay =t 2-11 = 12--

Where l and d are the distance from intersection stop bar to upstream RFID detector and downstream RFID detector, um is the saturation flow speed. But because of the queue in intersection, the vehicle will be forced to

decelerate and accelerate in the link (shown as point Bi and B2 in Fig.3), so the real delay stay in queue should be described by following equation:

t delay = Delay - tad (2)

Where tad is the sum of acceleration and deceleration delay. We assume that the vehicle speed in point A is approximately equal to the saturation flow speed for the detector is near to the intersection.

Fig.3 Shock wave and queue propagation

The shockwave profile on intersection is also showed in Fig.3. The traditional Lighthill-Whitham-Richards (LWR) traffic flow model hypothesizes that flow is a function of density at any point of the road. A shockwave is defined as "the motion (or propagation) of an abrupt change in concentration", for better explanation , we simply assume that queue has been fully discharged before the end of green phase. At the beginning of the red phase on intersection, vehicles from the upstream are forced to stop, which creates a queuing shockwave (wi in Fig.3) propagating to upstream. At the beginning of the effective green on intersection, vehicles begin to discharge at the saturation flow rate (assume there is no blockage downstream) forming the discharge shockwave (w2 in Fig.3),

which propagating upstream from the stop-line. The discharge shockwave usually has a higher speed than queuing shockwave, so the two waves will meet some time after the start of the green, which is the time that the maximum queue length is reached (indicated as point Q in Fig.3). The triangle ODQ shows the relationship of delay and queue length. The shockwave motion described above will repeat from cycle to cycle.

Traffic shockwave theory is derived from LWR model when applying the method of characteristics to analytically solve the partial differential equation in the model. Basically, a shockwave velocity can be determined by following equation:

w Aq q 2 - qj k 2 u 2-kiui

w = — =-=--(3)

Ak k2-ki k2-ki

Where q^, ki are foe flow and density of the upstream traffic and q2 , k2 are the fl°w and density of the downstream traffic.

At the beginning of the effective green, vehicles begin to discharge at saturation flow rate (assume there is no congestion downstream) forming the shock wave which is defined as discharge shockwave w2 in Fig. 3 at the stop line moving upstream with speed:

, 1 m_

km - k j

kmum-0 _ kmum

km - k j

km - k j

where um is the saturation flow speed, kj is the jammed flow density. Vehicles begin to discharge at the saturation flow rate start from stop line when it turns green, suppose their speed was um and density was km . As we know, km is less than kj, so w2 is negative and that means the shockwave's direction is opposite to the movement of traffic flow.

The model proposed here is aimed to estimate the maximum queue length in a cycle. As mentioned above, time t2 and t0 can be recorded by the RFID detector, wave w2 has a constant velocity (as shown in Eq. (3) if we

assume saturation flow rate qm , saturation density km , and jam density kj are known a priori), so in this

research, the delay time tdelay and wave w2 are utilized to identify the coordinate of point Q, i.e., the location

and time when maximum queue is reached, then the maximum queue length can be known. In this model, we assume that there are no random events on the road, and that queue has been fully discharged before the end of green phase, so the delay is only related to signal control and queue.

As shown in Fig.4, the red time is tred , line DQ is determined by the wave w2 and tred , line C2A is determined by t2 and um , so we can identify the coordinate of point A by line DQ and C2A, i.e., the location and time when the car begins to drive, the segment length of AB is the delay time tdeiay, so the coordinate of point B

also can be known, which is the location and time the car meets queue, line OB can be identified at the same time. The line OB and DA will meet at point Q, which is the time that the maximum queue length is reached, so the maximum queue length I, can be known at last.

■ Distance Green

Red RFID detector - •

CI \ tdelay Fig.4 Solution of model

From the description above, the maximum queue length in a cycle based on the queue delay of one car can be calculated by the following equation:

W2(u m 12- d )(u m tred - um t2

(Um - W2)(Umt 2- Umt delay - d)

In fact, there will be a number of cars with RFID tag pass through the intersection during a cycle, so we can use the average length to estimate the real queue length, the queue length can be determined by following equation:

4. Implementation

The intersection of Jin Xiang-he road and Beijing dong road in Nanjing is selected as the test site, whose location and detector layout are shown in Figure 5. In the test, two RFID detectors are used to detect the cars who has RFID tag and record the time the car passed by. The test last from 7:00am to 10:00am, August i, 20ii, during the test, the intersection is under timing control, the cycle is 180s and tred is 75s.

Fig. 5 Test site

During the test, we record the real queue length and the estimate queue length in each cycle, the results are shown in Fig. 6, in which we can know the profile of them. As shown in Figure 6, the estimate queue length is very close to the real queue length in most of the time.

Queue Length(m)

-Estimate Length Real Length

7:00:00 7:27:00 7:54:00 8:21:00 8:48:00 9:15:00 9:42:00Time(s)

Fig. 6 Test results of queue length

To evaluate the performance of the results, the relative error of estimate queue length in each cycle is shown in Fig. 7, the relative error ranges from 0 to 22% and it is less than 15% in most of the cycles.

Fig. 7 The relative error of estimate queue length

The distribution of relative error is shown in Tab.2, from which we can find more details. As we can see in Tab.2, about 92% of the relative error is smaller than 15%, so we can conclude that the result is acceptable in application.

Tab.2 The profile of relative error

Grouping of relative error 0~5% 5%~10% 10%~15% >15%

Proportion 9% 30% 53% 8%

5. Conclusions

In this paper, we propose a model to estimate the real-time queue length at signalized intersections, the model solves the problem of measuring intersection queue length by exploiting the queue delay of individual vehicles instead of counting arrival traffic flow in the signal cycle. The RFID system in Nanjing is used to test the algorithm, ow field-test results from an intersection in Nanjing demonstrate that the algorithms are very effective in estimate the maximum queue length in a cycle. We should note that, although the algorithms is effective enough on a single approach of a signalized intersection during the morning peak from 7:00 to 10:00, but maybe it's not easily to expand it to an arterial, or a network of intersections, and the effectiveness of this model in the situation of low saturation degree is still need to be verified. This research is expected to contribute to the future research on the queue length estimation of signalized intersection, as mentioned before, different types of intersection may require different estimate strategies, so future research may focus on how to improve the algorithm based on RFID system.

Acknowledgements

This research is financially supported by the National High-tech R&D Program (863 Program) of China (Project No. 2012AA112306).

References

Webster, F. (1958). Traffic signal settings. Road Research Technical Paper 39. Road Research Laboratory, Her Majesty's Stationery Office, London.

Newell, G.F. (1965). Approximation methods for queues with application to the fixed-cycle traffic light. SIAMReview 7. Gazis, D.C. (1974). Traffic Science. Wiley Interscience, New York.

Sharma, A., Bullock, D.M., Bonneson, J., (2007). Input-output and hybrid techniques for real-time prediction of delay and maximum queue length at a signalized intersection. Transportation Research Record, #2035, TRB, National Research Council, Washington, DC, pp. 88-96. Vigos, G., Papageorgiou, M., Wang, Y., 2008. Real-time estimation of vehicle-count within signalized links. Transportation Research, Part C 16 (1), 18-35.

Lighthill, M.J., Whitham, G.B. (1955). On kinematic waves I Flood movement in long rivers. II A theory of traffic flow on long crowded

roads. Proceedings of Royal Society (London) A229, 281-345.

Michalopoulos, P.G., Stephanopoulos, G. (1981) An apphcation of shock wave theory to traffic signal control. Transportation Research,

Part B 15, 31—51.

Stephanopoulos, G., Michalopoulos, P.G. (1979). Modelling and analysis of taffic queue dynamics at signalized intersections. Transportation Research, 13A.

Henry X. LIU, Xinkai WU, Wenteng MA, et al. (2009). Real-time Queue Length Estimation for Congested Signalized Intersections.

Transportation Research Part C, 17, 412-427.

Xinkai Wu, Henry X. Liu, Douglas Gettman. (2010). Identification of Oversaturated Intersections using High-resolution Traffic Signal Data.

Transportation Research Part C, 18,626-638.

WANG Dianhai,JING Chunguang,QU Zhaowei. (2002). Application of Traffic-wave Theory in Intersections Traffic Flow Analysis. China Journal of Highway and Transport, 15(1) Jan 93-96.

FAN Hongzhe. (2007). Research on Control Methods of Signalized Intersections under Queue Length Restriction. Chang chun, Jilin University.

RFID Passive Tags Track Cars (2004). <http://www.radio.weblogs.com/0105910/2004/08/10.html>.

W. Wen. (2010). An intelligent traffic management expert system with RFID technology. Expert Systems with Applications. 37 (2010) 3024— 3035.