Methodology for Variable Speed Limit Activation in Active ... - Core

Available online at www.sciencedirect.com

ScienceDirect Procedia - Social and Behavioral Sciences 96 (2013) 2129 – 2137

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

Methodology for Variable Speed Limit Activation in Active Traffic Management GU Shao-longa,*, MA Juna, WANG Jun-lia, SUI Xiao-qinga, LIU Yana a

Department of Traffic Management Engineering, People s Public Security University of China, Beijing 102623, China

Abstract

Active Traffic Management (ATM) is a tool that can stabilize traffic flow and improve safety by dynamically managing and controlling traffic based on the prevailing traffic conditions. This paper analyzes the traffic flow characteristics of G6 freeway in Beijing of China based on survey data. The relationships among traffic parameters such as standard deviation of speeds (SDS), average speed and occupancy are established in order to calibrate the critical thresholds for the activation of a VSL system in ATM. Furthermore, the control logic of a VSL system is proposed based on the relationships between SDS and other parameters. © Published by Elsevier Ltd. Open © 2013 2013The TheAuthors. Authors. Published by Elsevier B.V.access under CC BY-NC-ND license. Selection peer-review under under responsibility of Chinese Association (COTA). Selectionand and/or peer-review responsibility of Overseas Chinese Transportation Overseas Transportation Association (COTA). Keywords: Variable Speed Limit; Active Traffic Management; traffic flow characteristics; control logic; standard deviation of speeds.

1. Introduction The successful implementation of ATM in European countries and the United States provides new ideas and methods for the control and management of freeway systems in China. As a part of ATM system, VSL is an important control strategy which plays a positive role in stabilizing the traffic flow, preventing or postponing a breakdown and increasing the efficiency of freeways. Although there have been several researches of VSL in China, it is quite different for VSL to be a constituent part of ATM than to be used alone in terms of control strategy and scheme. At the same time, there are few research works about how to calibrate the critical thresholds for VSL activation from the view of applications in ATM system, which would be the crucial issues for the implementation and popularization of ATM technology in China. * Corresponding author. Tel.: +86-10-83906251; fax: +86-10-83906338. Supported by the Project of Capital Social Security Research Base (No. 11JDZR05). Supported by the Project of National Science & Technology Support Program (No.2013BAG18B01). E-mail address: [email protected]

1877-0428 © 2013 The Authors. Published by Elsevier Ltd. Open access under CC BY-NC-ND license. Selection and peer-review under responsibility of Chinese Overseas Transportation Association (COTA). doi:10.1016/j.sbspro.2013.08.240

2130

Gu Shao-long et al. / Procedia - Social and Behavioral Sciences 96 (2013) 2129 – 2137

Based on the data collected from a Beijing suburban freeway, this paper introduces the concept of SDS which is an indicator for speed synchronization and traffic stability and conducts an analysis on relationships among average speed, occupancy and SDS to calibrate the critical thresholds, and proposes a method for VSL activation under the prevailing conditions of the Beijing suburban freeway, which is expected to provide the foundation and methodology for the future implementation of ATM in China. 2. Literature review The development of ATM technology came along with the application and improvement of control strategies such as VSL, ramp metering and shoulder management. The applications of VSL system were synchronous with or even earlier than the theoretical study and the evaluation of effectiveness. With the development of ATM technology, the control strategies have been constantly changing within the process of the VSL application from weather response and work zone response to congestion and incident management. Allaby (2006) introduced several cases of existing VSL systems deployed in European countries and the United States, and summarized the control strategies of different VSL systems. In London, VSL system on M25 was used to delay the onset of flow breakdown and aid flow recovery by harmonizing speeds and reducing the severity of shockwaves. Real-time traffic data on M25 were collected through loop detectors and the activation of VSL depended on the measure of traffic volumes. When station volumes reached 1650 vphpl, the speed limits varied from 60 mph to 70 mph; and when the volumes reached 2050 vphpl, the speed limit further reduced to 50 mph. In Netherlands, VSL system on A2 also used loop detectors to collect the data of speed and volume. When station volumes approached capacity, the displayed speed limit could reduce to 90 km/h or 70 km/h from the original limit of 120 km/h according to the average speed detected. This system aimed to reduce the speed difference within and between lanes. In most cases, rule-based activation control schemes were applied as the VSL control strategy. Traffic flow parameters such as volume, speed and density were chosen as the critical indicators to decide the activation time of VSL via predetermined control logic. Although rule-based VSL control logic aims at preventing congestions actively, the values of traffic flow parameters are different from road to road. Hence the rule-based VSL activation control schemes cannot be applied to other freeways directly. Different from the control logic above, Papageorgiou, et al. (2008) considered that it was more stable and effective to control when applying the actual changes of slope of the occupancy-flow curve as an indicator for VSL activation, but the specific control algorithm was not given. Kühne (1991) proposed a VSL activation combination control strategy which applied actual traffic density, average speed and SDS as three thresholds. The control effect on speed harmonization is obvious. The earlier studies of freeway VSL control in China mainly focused on the independent use of VSL to control traffic flow. Factors such as density, pavement condition and weather condition had been taken into consideration as the thresholds for VSL activation. Besides, intelligent algorithms were also used to control VSL system, e.g. support vector machine algorithm (Liang et al., 2005), fuzzy neural network algorithm (Chen, 2009), genetic algorithm and so on. These control algorithms were mainly used to determine the speed limits under adverse weather and pavement conditions in order to improve traffic safety on freeways, which were different from the active control concept of ATM. He, et al. (2008) analyzed the traffic flow stability and synchronization among lanes around critical density to obtain the critical parameters for the identification of traffic phases based on empirical data of Beijing 2nd Ring Road. Probabilities of speed transition and deviations of traffic speeds among lanes were studied in detail. Lin et al. (2007) and Huang et al. (2010) combined the mainline VSL control with ramp metering and proposed a fuzzy-based control algorithm using current traffic flow density and on-ramp queue length as activation thresholds.


3. Methodology for VSL activation in ATM system In this paper, traffic flow data collected from 50 stations on G6 freeway near Beijing suburban are analyzed. Based on the historical data, there are high probabilities that traffic breakdowns or congestions happen on station 51032. Hence station 51032 is chosen as the study site. Two-minute interval data including traffic volume, average speed and occupancy of each lane from April 29, 2011 to May 6, 2011 collected by microwave detectors are analyzed in detail. Since the traffic density cannot be obtained directly and the converting occupancy to density may cause error, occupancy is used instead of traffic density. 3.1. Traffic flow characteristics on G6 freeway As shown in Fig. 1, the traffic flow parameters (i.e., flow rate, average speed and occupancy) and their relationships on station 51032 of G6 freeway demonstrate the following features: Obvious distinction of traffic states exists among three lanes in G6 freeway. The median lane represented by blue color has the highest throughput; the shoulder lane represented by red color has the lowest and the middle lane is in between. The driving habits of Chinese drivers on freeways would be the main reason. Median lanes become to be the first choice for drivers, especially for new drivers, to avoid the interruption of vehicles from on- or exit-ramps. Capacity drop between free-flow capacity and queue discharge rate on G6 freeway is obvious. Once congested due to an incident, its impact on the duration time of congestion is significant. We can observe from the speed-flow diagram that the maximum flow rate will reaches to 1800 vph when speed reduces from free flow to 65 km/h. When speed increases from congested flow to 20 km/h, the maximum flow rate can only be at the magnitude of 1400 vph which is 400 vph less than free-flow capacity. When speed lies between 20 km/h and 65 km/h, traffic flow demonstrates obviously instability and the weak correlation between flow rate and speed. From the flow rate-occupancy diagram, we can observe that there is an obvious linear relationship between flow rate and occupancy when occupancy is lower than 20%, and the traffic is in free flow condition. Traffic speed gets lower and congestion forms gradually when occupancy is higher than 40%, and the linear relationship between volume and occupancy is not obvious. The traffic is in the state of unstable when the occupancy varies from 20% to 40%, and the correlation between volume and occupancy are weak. 120 100

Speed (km/h)

80 60 40 20 0 0

300

600

900

1200

Flow rate (vph)

(a)

Speed-flow relationship

1500

1800

2131

2132

Gu Shao-long et al. / Procedia - Social and Behavioral Sciences 96 (2013) 2129 – 2137 2000 1600

Flow rate (vph)

1200 800 400 0 0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

80.0

90.0 100.0

70.0

80.0

90.0

Occupancy (%) (b) Flow-occupancy relationship

120

Speed (km/h)

100 80 60 40 20 0 0.0

10.0

20.0

30.0

40.0

50.0

60.0

100.0

Occupancy (%) (c) Speed-occupancy relationship Fig. 1.Fundamental diagrams on station 51032 of G6 freeway for three lanes (RED -shoulder lane, YELLOW-middle lane, BLUE-median lane)

The interpretation of results analysis of traffic flow characteristics provides an essential basis to calibrate the thresholds for VSL activation. Moreover, the determination of capacity, critical speed and critical occupancy is of vital importance to the effect of VSL control in ATM. However, the traffic flow characteristics of freeways in China are different from road to road. What s more, since the distribution of unstable traffic state points in fundamental diagrams is irregular, it would be difficult to determine critical parameters (especially critical speed and critical occupancy) from Fig. 1 directly. The objective of VSL in ATM is not to reduce the average speed of traffic flows, but to decrease the speed variance among vehicles and to stabilize traffic flows eventually. Kühne (1991) firstly introduced SDS as the parameter for VSL activation. Good results had been achieved in speed harmonization. Therefore, the analysis on SDS is used to determine the VSL activation method in this paper. VSL system control should be a dynamic process which varies with the real-time data collected by the detectors compared with the study on identification of traffic phases (He, et al., 2008). The deviations of traffic speeds among lanes cannot reflect the tendency of speed variation with time and the relationship between traffic


flow and speed variation with time. Hence, SDS, average speed and occupancy of single lane in 10 minutes, i.e. 5 detecting intervals, are analyzed in this paper. Because the detected speeds are the average speeds within 2-minute interval, the value of SDS calculated may be lower than the reality. Median lanes are chosen as the study object to get more information of different phases with regard to the habit of drivers lane choice in China. 3.2. Critical parameters for VSL activation 3.2.1. Relationship between average speed and SDS Observed from the distribution of scattered points (see Fig. 2), the magnitude of SDS is concentrated on relatively low values when average speed is lower than 20 km/h or higher than 65 km/h. The traffic flow under this situation always is in the state of stability. When average speed lies between 20 km/h and 65 km/h, the distribution of SDS is dispersive and its values are relatively high, and corresponding traffic flow becomes unstable. This result is in accordance with the conclusions made from the fundamental diagrams of basic parameters. 35.0 30.0 SDS km/h

25.0 20.0 15.0 10.0 5.0 0.0 0

10

20

30

40 50 60 Average speed km/h

70

80

90

100

80

90

100

Fig. 2.Scatter diagram of SDS and average speed

35.0 30.0 SDS km/h

25.0 20.0 15.0 10.0 5.0 0.0 0

10

20

30

40 50 60 Average speed km/h

70

Fig. 3.Time-dependent curve of SDS and average speed

As shown in Fig. 3, the process of speed transition is clear and obvious. The relation curve between SDS and average speed appears parabolic when the traffic flow converts from free flow to stable low-speed congested flow. Furthermore, SDS reaches its peak value when average speed is approximately between 40 km/h and 50

2133

2134


km/h, traffic flows in these cases are on the border of unstable operations. The regression analysis of scattered points demonstrates that SDS is lower than 10 km/h and the possibilities of speed transition is nearly zero when average speed is higher than 76 km/h (see Fig. 4). When average speed lies between 68 km/h and 76 km/h, SDS decreases with the increase of average speed. In addition, the traffic flow is unstable and speed transition is much easier to happen though SDS is less than 5 km/h. SDS gets its maximum when average speed reaches to 43 km/h, which is taken as the value of critical speed. 30.0 2

y = -0.0214x + 1.8394x - 22.566

25.0

y = 0.7394e

2

R = 0.8282

0.0596x

SDS (km/h)

20.0 15.0 y = 0.0925e

10.0 5.0

0.0456x

y = -0.1915x + 17.226

0.0 0

10

20

30

40

50

60

70

80

90

100

Average speed (km/h)

Fig. 4.Fitting curves of SDS and average speed

3.2.2. Relationship between occupancy and SDS The scatter diagram of SDS and occupancy shows that the relationship between occupancy and SDS is similar to that of average speed and SDS (see Fig. 5). 35.0 30.0

SDS km/h

25.0 20.0 15.0 10.0 5.0 0.0 0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

80.0

Occupancy % Fig. 5.Scatter diagram of SDS and occupancy

As shown in Fig. 6 and Fig. 7, the traffic flow is in stable state when occupancy is below 9%. Moreover, SDS decreases with the increase of occupancy and speed transition is almost impossible. When occupancy is above 47%, SDS basically maintains at the level of below 5 km/h, and the traffic breakdown would occur with higher occupancy. When occupancy lies from 9% to 13%, SDS maintains at the level of 5 km/h. In this case, the traffic flow is unstable and speed transition is much easier to happen though SDS is below 5 km/h. When occupancy is about 29.9%, SDS reaches to its maximum and that is named as critical occupancy which corresponds to the critical speed mentioned above.


2135

35.0 30.0 SDS km/h

25.0 20.0 15.0 10.0 5.0 0.0 0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

80.0

Occupancy % Fig. 6.Time-dependent curve of SDS and occupancy

30

y = -0.0006x 3 + 0.0043x 2 + 1.3619x - 11.129 R2 = 0.5862

SDS (km/h)

25 20 15 10

y = -1.6248Ln(x) + 8.5841

5 0 0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

80.0

Occupancy (%) Fig. 7.Fitting curve of SDS and occupancy

3.3. The control logic of VSL activation SDS studied in this paper is used to measure stability and synchronization of speed in single lane within certain period of time. Analysis on the relation curves of average speed, average occupancy and SDS demonstrates that the speed synchronization of continuous flow can reflect the stability of traffic flow. Therefore, decreasing speed variances and sustaining speed synchronization of high-speed and low-occupancy traffic flow have active effects to ensure safety, smoothness and stability of freeway traffic flow without external disturbance. Based on the relation curves of average speed, average occupancy and SDS, the freeway flow shows unstable feature when speed reduces from 76 km/h to 68 km/h and occupancy increases from 9% to 13%. To restrain the difference of lane speeds and prevent the speed transition, isolated speed limits control with high-value (i.e. 70 km/h or 75 km/h) can be used when the traffic is ready to changes into unstable phase. Observed from the analysis of traffic data on G6 freeway, the traffic speed always goes in a similar trend that reduces from 80 km/h to 60 km/h at the beginning of morning peak (from 6:00 am. to 8:00 am.). Hence VSL system should be activated when the average speed detected reduces to threshold value (76 km/h) during the morning peak hours. This is in accord with the positive and preventive effect of ATM strategies. The freeway traffic is stable in most cases (including low-occupancy and high-occupancy) as depicted both from the fundamental diagrams and the relation curves of average speed, average occupancy and SDS. In other

2136


words, the duration of unstable phase is always short and temporary, but it is just the time when VSL control works. In terms of speed synchronization, the critical speed and critical occupancy are of vital importance. When speed or occupancy reaches or approaches the thresholds, synchronization of speeds gets the worst and the stability of traffic flow comes to the lowest. Furthermore, it is very possible to induce traffic breakdown and finally lead to congestion. At the beginning of phase transition, isolated or coordinated speed limit control with low-value (40 km/h to 50 km/h) can be implemented as early as possible before the traffic parameters reach the critical thresholds (critical speed, critical occupancy and extreme value of standard deviation). VSL can reduce the speed variance of unstable traffic flow, and stabilize traffic flow to postpone the onset of congestion or minimize the impact of congestion. Generally speaking, there is high randomness on speed transitions which happen when traffic flow is unstable, but the formation of congestion always comes along with the process that speed, occupancy and SDS approach or move away from critical values (see Fig. 3 and Fig. 5). Based on the law of traffic phase transition above, we can determine thresholds for VSL activation according to the changes of distance or slope between detected real-time values and critical values of speed or SDS. 4. Conclusion In this paper, the relationships between traffic flow parameters (speed and occupancy) and SDS are established. The critical values of speed, occupancy and SDS as thresholds are calibrated for activation of VSL control based on survey data. These values cannot be simply determined from the fundamental diagrams of traffic flow basic parameters. Based on the empirical data of G6 freeway in Beijing, two types of methods for VSL activation are proposed: one is the isolated speed limits control method with high-value (i.e. 70 km/h or 75 km/h) when the freeway traffic comes into unstable phase at the beginning of morning peak (from 6:00 am. to 8:00 am.); another is the isolated or coordinated speed limits control method with low-value (40 km/h to 50 km/h) at the beginning of phase transition. To ensure the effect of the second method, VSL system should be activated as early as possible before the traffic parameters reach to their critical values. Furthermore, accurate threshold values need to be determined according to the changes of distance or slope between real-time detected values and critical values of speed or SDS, which needs further studies in the future. References Allaby, P. E. (2006). An evaluation of the safety and operational impacts of a candidate variable speed limit control strategy on an urban freeway. The University of Waterloo. Breton, P., Hegyi, A., & Schutter, B. De. (2002). Shock wave elimination/reduction by optimal coordination of variable speed limits. In: . (pp.197 220). Singapore. Proceedings of the IEEE 5th International Conference on Brinckerhoff, P. (2010). Synthesis of active traffic management experiences in Europe and the United States. Publication Number FHWAHOP-10-031. Chen, D. S. (2009). Study on highway mainline speed limit control. Master s thesis. Xi an: . Chen, J. Y. (1993). Function and control of highway variable speed sign. Journal of Tongji University, 9(3): 388 - 391. Hegyi, A., Schutter, B. De., & Hellendoorn, J. (2002). Optimal coordination of variable speed limits to suppress shock waves. In: Proceedings of the 7th TRAIL Congress. (pp. 197 220). Rotterdam, Netherlands. Hegyi, A. (2004). Model predictive control for integrating traffic control measures. TRAIL Thesis Series, Delft University Press, The Netherlands. He, S. Y., & Guan, W. (2008). Empirical analysis of phase transitions of traffic flow at urban expressway. China Journal of Highway and Transport, 21(5): 81 - 86. Huang, C. P., Jiang, M., & Chai, G. (2010). Fuzzy control for ramp metering and variable speed limitation of freeway. Computer Technology and Development, 12(12): 38 41. Kühne, R. D. (1991). Freeway control using a dynamic traffic flow model and vehicle re-identification techniques. Transportation Research Record, 1320: 251 - 259.


2137

Liang, X. R., et al. (2005). Control speed limitation on freeway based on support vector machine. Computer Engineering and Applications, 41(34): 178 - 180. Lin, S. W., & Lin, Y. (2007). Urban expressway control based on fuzzy logic and variable speed limit. Journal of Transportation Systems Engineering and Information Technology, 2(1): 50 - 56. Mirshahi, M., et al. (2007). Active traffic management: the next step in congestion management. Federal Highway Administration. Nissan, A. (2010). Evaluation of variable speed limits: empirical evidence and simulation analysis of Stockholm's motorway control system. Doctoral dissertation. Royal institute of Technology. Stockholm, Sweden. Papageorgiou, M., Kosmatopoulos, E. &Papamichail, I. (2008). Effects of variable limits on motorway traffic flow. In: Proceedings of the 87thAnnual Meeting of the Transportation Research Board. Washington, DC. Sisiopiku, V., et al. (2009). Implementing active traffic management strategies in the U.S. UTCA Report Number 08206. Smulders, S. (1990).Control of freeway traffic flow by variable speed signs. Transportation Research Part B, 24B(2): 111 132.