Available online at www.sciencedirect.com Available online at www.sciencedirect.com Available online at www.sciencedirect.com
ScienceDirect ScienceDirect Transportation Research Procedia 00 (2016) 000–000 ScienceDirect
Transportation Research Procedia 00 (2016) 000–000 Transportation Research Procedia 22 (2017) 607–614
www.elsevier.com/locate/procedia www.elsevier.com/locate/procedia www.elsevier.com/locate/procedia
19th EURO Working Group on Transportation Meeting, EWGT2016, 5-7 September 2016, 19th EURO Working Group on Transportation Meeting, EWGT2016, 5-7 September 2016, Istanbul, Turkey Istanbul, Turkey
Simulation-based Variable Speed Limit Systems Modelling: An Simulation-based Variable Speed Limit Systems Modelling: An Overview and A Case Study on Istanbul Freeways Overview and A Case Study on Istanbul Freeways Mohd Sadat*aa, Hilmi Berk Celikogluaa Mohd Sadat* , Hilmi Berk Celikoglu a a
Istanbul Technical University,Istanbul 34469,Turkey Istanbul Technical University,Istanbul 34469,Turkey
Abstract Abstract Modern transportation systems aim at maximizing the use of available resources in a sustainable manner to deliver efficient and Modern transportation systems aim Speed at maximizing the use of available in a sustainable to improve deliver efficient safe movement of traffic. Variable Limit (VSL) system is one ofresources the techniques adopted inmanner order to mobility.and In safe movement of traffic. Variable Speed Limit (VSL) system is one of the techniques adopted in order to improve mobility. this study, we analyse this system using simulation techniques on a 5.2 kilometre section of Istanbul Freeway D100. Being one In of this study,congested we analyse this in system using Istanbul simulation techniques on a an 5.2excellent kilometreopportunity section of Istanbul Freeway D100. Beingofone of the most cities the world, freeways provide to test the potential benefits VSL the most Latest congested cities in thealong world, Istanbul freeways provide to test theare potential benefits VSL systems. advancements with comprehensive literatureanofexcellent this fieldopportunity based on simulation included in thisofstudy. systems. Latest advancements withVISSIM comprehensive literature this field to based on simulation are included thisvolume, study. Microscopic traffic simulation along software is used along withofMATLAB implement VSL algorithm basedinon Microscopic traffic simulation softwareTraffic VISSIM is used along MATLAB implement VSL algorithm basedwhich on volume, occupancy and average speed. Remote Microwave Sensorwith (RTMS) data isto provided by Istanbul municipality is used occupancy average speed. Remote Traffic Microwave Sensorwere (RTMS) data is for provided by Istanbul which used to calibrateand VISSIM.Scenarios with and without VSL system simulated morning hours. Itmunicipality is concluded that isdriver to calibrate to VISSIM.Scenarios with factor and without VSL system were simulated for morning hours. It is(CL) concluded driver compliance VSLs is an important for better results. Although 100% driver Compliance Level for VSLthat results in compliance VSLsimprovement is an important factor for better results. 100% driver Level (CL) therefore for VSL results significantlytohigher in performance compared to Although lower compliance, it is Compliance not a practical approach results in at significantly in performance compliance, it is is not a practical therefore results at 75% and 50%higher CL hasimprovement been discussed in this study.compared Evaluationtooflower network performance done in termsapproach of Total Travel Time (TTT) 75% and 50% CLwith has been discussed study. Evaluation of network performance is done in terms of Total Time (TTT) in network along volume, speed in andthis occupancy. Results show reduction in TTT and occupancy level alongTravel with improvement in along with volume, Fuel speedconsumption and occupancy. show reduction and occupancy level along with improvement in network average speed and volume. andResults emissions is also foundintoTTT be reduced in the network indicating sustainable in speed and volume. Fuel consumption and emissions is also found to be reduced in the network indicating sustainable andaverage environmental friendly mobility. and environmental friendly mobility. © 2016 The Authors. Published by Elsevier B.V. © 2017 Published by Elsevier B.V. B.V. © 2016The TheAuthors. Authors. Published by Elsevier Peer-review under responsibility ofScientific the Scientific Committee of EWGT2016. Peer-review under responsibility of the Committee of EWGT2016. Peer-review under responsibility of the Scientific Committee of EWGT2016. Keywords: Variable Speed Limits systems, Congestion, Compliance Level Keywords: Variable Speed Limits systems, Congestion, Compliance Level
* Corresponding author. Tel.: +905386537140 * Corresponding Tel.: +905386537140 E-mail address:author.
[email protected] E-mail address:
[email protected] 2214-241X © 2016 The Authors. Published by Elsevier B.V. 2214-241X 2016responsibility The Authors.of Published by Elsevier B.V. of EWGT2016. Peer-review© under the Scientific Committee Peer-review under responsibility of the Scientific Committee of EWGT2016. 2214-241X © 2017 The Authors. Published by Elsevier B.V. Peer-review under responsibility of the Scientific Committee of EWGT2016. 10.1016/j.trpro.2017.03.051
608 2
Mohd Sadat et al. / Transportation Research Procedia 22 (2017) 607–614 Author name / Transportation Research Procedia 00 (2016) 000–000
1. Introduction Growing traffic on freeways in modern times have created a need for a system which can manage traffic efficiently without expanding the infrastructure. As the world moves towards sustainable development, using resources efficiently is the need of the hour. Modern transportation systems aim at maximizing the use of available resources in a sustainable manner to deliver efficient and safe movement. Freeways were developed to provide high speed non-stop routes for vehicular traffic by separating conflicting traffic streams and limiting access to freeway. Traffic congestions on freeways cause delay, waste of fossils and safety hazards. Intelligent Transportation Systems (ITS) are used in modern transport facilities to minimize these problems. One of the techniques is Variable Speed Limit (VSL) systems in which speed of traffic stream is decreased in small values in order to harmonize the traffic flow before traffic breakdowns. There are various control algorithms which further optimize the performance of VSL system. Microscopic traffic simulation software provide an excellent platform to develop, test and improve the control strategies. The following sections discuss works on VSL systems and microscopic simulations. The effects of VSL system operation on traffic stream is also discussed. Methodology and control strategy is discussed in section 3.Finally a case study on Istanbul freeway network is completed in the last section. Benefits of VSL systems such as improvement in average volume and average speed along with reduction in occupancy and Total Travel Time (TTT) is presented along with reduction in emissions and fuel consumptions. Comparison of results for VSL system is done with 50% and 75% driver compliance level. 2. Literature review VSL techniques for freeway management have already been implemented in countries like Netherlands, UK, Denmark, and Australia. It is found to have lowered the level of congestion and emissions. There are several VSL algorithms used on freeways which are mainly based on variables of occupancy, speed and volumes. Critical values are defined in control algorithm at which VSL system actuates. In order to reduce the propagation of shockwaves during congestion, the desired speed is reduced upstream at critical values. METANET simulation models were developed for the VSL system (Hegyi et al., 2005) but were not implemented. Algorithms based on average occupancy thresholds are implemented on I-4 in Orlando, Florida while flow based algorithms were implemented on M25 Motorway in England and E6 motorway in Mölndal, Sweden.Studies have shown that VSL systems have helped in reducing Total Travel Time (TTT) in network and smoothened traffic flow (Abdel-Aty et al. 2006).Total Time Spent (TTS) in network was found to be 17.4% lower when proposed VSL was applied in coordination with ramp metering in a hypothetical section by Hegyi et al.(2005). Benefits of VSL were also documented by Papageorgiou et al. (2008). Safety levels were found to increase with VSL techniques on steep bottleneck of mountainous sections using microscopic simulation software VISSIM for simulation and calibration(Yu and Abdel-Aty 2014).Safety benefits of VSL were also realized by developing a crash model (Abdel-Aty et al.2006). Reduction of secondary collision in low visibility weather conditions was shown by Li et al. (2014) using a modified car following model. Co-operative VSL system (C-VSLS) is a strategy which decreases the involvement of the driver by directly communicating the speed to the vehicle (Grumert et al. 2015). Grumert et al.(2015) used study open source microscopic simulation software SUMO was used with Python. Results were better than conventional VSL systems in terms of steady flow and emissions. Reduction in CO level was 2.66% better than conventional VSL system in Grumert et al. (2015). However, the results were subjected to penetration level. At 70% penetration level due to rapid lane changing behaviour of non C-VSLS vehicles when C-VSLS equipped vehicles decelerate congestion might develop. Similar studies were conducted by Cao et al.(2015) and Khondaker and Kattan (2015) by using micro simulation.Greenhouse emissions were minimized with use of Fuel Consumption-Aware Variable Speed Limit strategies (FC-VSL) (Liu et al.2012). Similar in approach with previous strategies, FC-VSL aims at harmonizing flow which results in lower fuel consumption thus decreasing carbon footprint. Khondaker and Kattan (2015) reported 14.8% reduction in average fuel consumption using VSL in connected environment. A study on the effect of VSL on traffic flow characteristic (Allaby et al.2007) concluded that VSL showed desirable results with moderate to high congestion while producing undesirable results sparingly congested scenarios. Allaby et al.(2007) made modifications in the algorithm and concluded that there were scenarios where VSL was not needed but actuated due to small disturbance in flow. This triggering of VSL did not allow the traffic flow to recover as it would have done in case of non VSL scenario. Modification by included increasing the threshold speed at which VSL kicks in and decreasing the number of VSL upstream of congestion point which get actuated. However, driver compliance level was taken same for VSL and non
Mohd Sadat et al. / Transportation Research Procedia 22 (2017) 607–614 Author name / Transportation Research Procedia 00 (2016) 000–000
609 3
VSL cases. Most of the works on VSL systems were developed and tested on traffic simulators (Allaby et al. 2007; Grumert et al. 2015; Yu and Abdel-Aty 2014; Khondaker and Kattan 2015; Hadiuzzaman et al. 2015).These simulation software have their own models which can be modified by changing some parameters like driver aggression. They also allow user to implement their own algorithms by allowing support for programming tools like C++, VBA, Python C# and MATLAB. Thus simulation based models allow researchers to perform complex studies through computer aid without actual field study. Driver’s compliance to posted speed limits is very important for VSL strategy to effective. Hadiuzzaman et al. (2015) conducted a study in which they disproved that VSL increase the travel time under high compliance level (CL). There was no study which could clearly determine whether the control strategy of VSL was responsible for delay or the compliance level of drivers. The travel time and collision probability were found to improve in the study. Therefore, the CL is an important condition which is subjected to the way authorities implement the VSL on freeways. Nissan and Koutsopoulosb (2011) showed that advisory VSL failed to show any significant improvement in traffic condition even months after installation. Long et al. (2012) conducted a study in Missouri, USA which indicates that satisfaction level of both users and enforcing authority personals was low. Even in case of mandatory VSL sign it was practically not possible to enforce the speed limits. This is also indicative of the fact that hypothetical simulation may not necessarily produce same results as success of VSL system is heavily dependent on its acceptance by users. Driver compliance level was found to be higher for higher speeds (Yu and Abdel-Aty et al. 2014). Bhowmick et al. (2011) concluded that even at low compliance level, VSL was effective showing lower speed variance. 3. Methodology VSL algorithm in the present study is adopted from Allaby et al. (2007). VISSIM’s default car following model is used for simulation. VISSIM uses stochastic, time-step based, microscopic model that treats driver vehicle units as basic entities. It is based on Wiedemann 74 and Wiedemann 99 (VISSIM 8 Manual, 2015) which define car following model for longitudinal movement and a rule-based algorithm for lateral vehicle movement. Control strategy is implemented by integrating MATLAB with VISSIM via Component Object Model (COM) interface. Traffic volume, speed and occupancy RTMS data for every two minutes was provided by Istanbul Municipality. Data from RTMS 301, 533 and 534 are used as input for simulation in VISSIM. Calibration is done by modifying parameters until speed profiles matched with the field data. Parameters for calibration includes lane changing behavior, car following behaviors, headway and lane changing distance. Real volumes and simulated volumes are used to calculate the GEH Statistics (UK Highways Agency). ( )
= (
−
)−
( ) ( )
/2
(1)
Where Mobs(n) represents the RTMS volumes and Msim(n) represents the simulated volumes obtained from the VISSIM simulation. For all locations 90% of GEH values are within the error equal or less than 5. According to UK Highways Agency's Design Manual for Roads and Bridges (DMRB) 85% volumes in the simulated model should have GEH less than 5 for accurate representation of real-field traffic flow.
Figure 1 .Speed profile after calibration at Detector 2
Mohd Sadat et al. / Transportation Research Procedia 22 (2017) 607–614 Author name / Transportation Research Procedia 00 (2016) 000–000
610 4
Table 1.GEH statistics for Detector 2 GEH statistics for Detector 2
Mean
Standard deviation
Maximum value
Minimum value
1.69
1.09
5.35
0.12
Therefore, VISSIM simulation for the section is representative of actual traffic flow. Figure 1 shows simulated and observed speed distributions which shows the validation of calibration.Table 1 shows the GEH values for detector 2.Volume, occupancy and speed data are retrieved during simulation by MATLAB in real time. Desired speed is determined by running program code and sending back to VISSIM as represented in Figure 2. This process is repeated and VSL is updated after each minute because of constraint that lower displayed speed should sustain for at least 1 minute which is discussed in section 3.2.
Figure 2 .Calculation of desired speed
3.1. Case study test segment Many segments over D100 freeway in Istanbul suffer from recurrent congestions (Celikoglu and Dell’Orco 2007; Celikoglu and Dell’Orco, 2008, Celikoglu 2013a; Celikoglu 2013b; Celikoglu, 2014; Celikoglu and Silgu 2016). The section considered for the simulation is 5.2 km west bound 3 lane D100 freeway connecting Zincirlikuyu to Okmeydani (Abuamer and Celikoglu, 2016). There are three on-ramps from Barbaros Boulvd, Buyukdere Street and Caglayan. There is an off-ramp towards Caglayan. Congestion is found on ramp from Buyukdere Street and main link merge. This section provides excellent environment for testing the VSL algorithm as there is recurrent congestion during morning hours. Evening hours is relatively less congested therefore VSL system evaluation was carried out for morning hours. Figure 3 shows the speed profile for morning and evening hours for a span of six hours. Presence of on and off-ramps before and after this section respectively make the situation more complex. A section can be divided into segments of 400-800 m (Papageorgiou et al., 2008). Figure 4 shows the section having four VSL signs V1, V2, V3 and V4 along with placed detectors. Rectangular boxes represent the detectors for respective VSL signs. Detectors are addressed with respective VSL sign numbers henceforth.
Figure 3.Morning and evening speed profile
Mohd Sadat et al. / Transportation Research Procedia 22 (2017) 607–614 Author name / Transportation Research Procedia 00 (2016) 000–000
611 5
Figure 4.Selected section for VSL, D100
3.2 System Control Strategy This study adapts algorithms from Allaby et al. (2007) which is based on volume, occupancy and average speed. It is based on a study of a freeway in Toronto, Canada. At the time of the study there was limited information on driver’s response to VSL systems. The present study, along with implementation of this algorithm, aims to include driver’s compliance to VSLs in a case study on Istanbul freeways. Different critical values of average speed, volume and occupancy are varied during simulation in order to determine most suitable critical values for the algorithm i.e. values at which network performance is most desirable.The most suitable algorithm is shown in Figure 5. Results of this algorithm have been discussed in section 4. Driver compliance level is varied for a realistic approach in simulation. VISSIM’s ‘desired speed distribution’ attribute allows to set different level of compliance by the drivers. For each vehicle composition type, predefined percentage of drivers drive below desired speed while others above it.This allows for a realistic approach in VSL system implementation. In this study 50% and 75% driver compliance levels are considered by using ‘desired speed decision’ attribute in VISSIM which is called from MATALB via COM .For 50% CL, 20% of drivers kept their speeds below desired speed while 30% drivers kept it above. Similarly for 75% CL, 10% drivers kept the speeds below desired speed while 15 % kept it above it. During the simulation, 25th January (Monday) morning is found to be most congested at Detector 2 before VSL system implementation therefore, is chosen for VSL implementation. Detectors are programmed to relay 60 seconds occupancy, average speed and volume data to MATLAB via COM. New desired speed is calculated in MATLAB and relayed back to the simulator for VSL system display. The displayed speed is subjected to the following constraints: Difference between displayed speeds of two consecutive VSL signs should not be more than 20 km/h (Allaby et al., 2007). Lower speeds should sustain for at least one minute (Khondaker et al., 2015). These constrains prevent sudden speed changes and allow the traffic to recover before the sign is changed to higher speed values. Some network performances are also evaluated including environmental impacts and congestion levels expressed in terms TTT. For calculating fuel consumption, CO and NOx emissions in section VISSIM’s default model is used. Different cutoff values are used for volume, occupancy and average speeds. Algorithm returning best results in terms of network performance is being presented.
Figure 5.Algorithm for VSL control [adapted and modified from Allaby et al., 2007]
6 612
Author name / Transportation Research Procedia 00 (2016) 000–000 Mohd Sadat et al. / Transportation Research Procedia 22 (2017) 607–614
4. Simulation and Findings Simulations with VSL systems are carried out from 9 am to 12 pm with 50% and 75% driver’s compliance levels. Warm up period of 4 minutes is taken during the simulation. Passenger Car Unit (PCU) is taken as default vehicle composition. Figure 6 shows pre-and post-VSL volumes for Detector 2. It shows 8.21% and 7.5% improvement in average volumes for 75% and 50% driver’s compliance levels respectively. Figure 7 shows average speed profile for pre-and postVSL. There is stop and go traffic condition after 10 am during pre-VSL simulation. It is found that post-VSL traffic slows down but recovered after 10:20 am for both driver’s compliance levels (CL). For 75% CL, recovery to higher speeds is better than 50% CL with exception from time 11:20 am to 11:50 am.
Figure 6.Comparison of Pre-VSL and Post-VSL volumes at Detector 2
Figure 7.Speed profile for Pre-VSL and Post-VSL scenarios at Detector 2
Figure 8.Comparison of Pre-VSL and Post-VSL occupancy at Detector 2
7
Author name / Transportation Research Procedia 00 (2016) 000–000
Mohd Sadat et al. / Transportation Research Procedia 22 (2017) 607–614
613
Figure 8 shows the occupancy levels which also shows improvement in post-VSL scenarios and slightly better results for 75 % CL. Low occupancy is necessary for better utilization of freeway capacity as after reaching the maximum capacity the volume breaks down and traffic flow decreases. It can be observed that traffic nearly broke down at three points during VSL implementation which corresponds to congestion periods of pre-VSL. Overall VSL system helped in bringing the occupancy down. Table 2 shows the average volumes, average speed, and average occupancy for the all detectors along Total Travel Time for pre-and post-VSL scenarios. It can be observed that average volume is improved by 8.58% for 75% CL and 5.75% for 50% CL. Similarly, average speed and occupancy shows improvement of 8.2% and 7.85% for 75% CL respectively. Total Travel Time (TTT) is decreased by 8.93% for 75 % and 7.4% for 50 % CL. TTT is calculated for the vehicles entering the network by main link and leaving it from the main link thus traversing distance of 5.2 kilometres.VISSIM’s default model is used for calculating TTT.It can be observed from the results that higher driver compliance level show better results in terms of traffic flow dynamics. Table 2.Results for Pre-VSL and Post –VSL scenarios for all detectors All detectors
Pre-VSL
Post –VSL (75% CL )
Post –VSL (50% CL )
Average volume (v/h/l)
1130
1227
1195
Average speed (km/h)
46
49.8
48.2
Average occupancy (%)
28
25.8
26.1
Total Travel Time (seconds)
526
479
487
Figure 9a shows the total fuel consumption of all vehicle passing through network for pre-VSL and post-VSL scenarios. There is 28.5% reduction in fuel consumption for 75% CL and 27% for 50% CL. Figure 9b shows reduction in CO and NOx emissions with VSL system compared to the Pre-VSL scenario. Proposed VSL system thus shows improvement in terms of enhanced mobility and environmental benefits.
Total Fuel Consumption (l) 2000
100
1620.27
Total emissions (kg)
CO
64.15
65.43
80
1500 1158.04
NOx
89.76
1181.19 60
1000 40 500
20
0
0 Pre-VSL
Post-VSL (75% Post-VSL (50% CL) CL)
17.46
Pre-VSL
12.48
12.73
Post-VSL (75% Post-VSL (50% CL) CL)
Figure 9.a) Total fuel consumption; b) total CO and NOx emissions.
5. Conclusion and Discussion This study provides an overview of simulation based VSL system modelling and investigates potential benefits of VSL on D100 Istanbul freeway from Zincirlikuyu towards Okmeydani.Volume, occupancy and speed based algorithms are implemented by integrating MATLAB with VISSIM via COM interface. VSL system is implemented for morning hours with 50% and 75% driver compliance levels. From the results it can be concluded that proposed VSL systems helped in minimizing the congestion levels even at moderate compliance levels. It can also be concluded
614 8
Mohd Sadat et al. / Transportation Research Procedia 22 (2017) 607–614 Author name / Transportation Research Procedia 00 (2016) 000–000
that increased compliance levels produce better desired results therefore mandatory speed limits should be enforced to minimize the congestion. It is also found that fuel consumption is significantly decreased along with reduced CO and NOx emissions thus shows promising environmental friendly aspect of proposed VSL system. In this study, onramps are assumed to be without control. In further studies, we will test an integrated traffic responsive ramp metering with the VSL system on the selected section of D100 freeway. References Abdel-Aty, M., Dilmore, J. and Dhindsa, A., 2006. Evaluation of variable speed limits for real-time freeway safety improvement. Accident analysis & prevention, 38(2), pp.335-345. Abuamer, I.M., Celikoglu, H.B., 2016. Local Ramp Metering Strategy ALINEA: Microscopic Simulation Based Evaluation Study on Istanbul Freeways. Transport Research Procedia – Proceedings of the 19th Euro Working Group on Transportation, EWGT 2016, 05-07 September 2016, Istanbul, Turkey. Allaby, P., Hellinga, B. and Bullock, M., 2007. Variable speed limits: Safety and operational impacts of a candidate control strategy for freeway applications. Intelligent Transportation Systems, IEEE Transactions on, 8(4), pp.671-680. Behrisch, M., Bieker, L., Erdmann, J. and Krajzewicz, D., 2011, October. Sumo–simulation of urban mobility. In The Third International Conference on Advances in System Simulation (SIMUL 2011), Barcelona, Spain. Bhowmick, A., Qiu, T.Z. and Hadiuzzaman, M., 2011. Driver compliance analysis of variable speed limit based freeway traffic control. In Proceedings of the 11 st International Conference of Chinese Transportation, Beijing: American Society of Civil Engineers (pp. 4181-4192). Cao, J., Hu, D., Luo, Y., Qiu, T.Z. and Ma, Z., 2015. Exploring the impact of a coordinated variable speed limit control on congestion distribution in freeway. Journal of Traffic and Transportation Engineering (English Edition), 2(3), pp.167-178. Celikoglu, H.B., Dell’Orco, M., 2007. Mesoscopic simulation of a dynamic link loading process. Transportation Research Part C: Emerging Technologies 15 (5), 329-344. Celikoglu, H.B., Dell’Orco, M., 2008. A Dynamic Model for Acceleration Behaviour Description in Congested Traffic. Proceedings of the 11th IEEE International Conference on Intelligent Transportation Systems (ITSC 2008) pp. 986-991, DOI: 10.1109/ITSC.2008.4732521. Celikoglu, H.B., 2013a. Flow-Based Freeway Travel-Time Estimation: A Comparative Evaluation Within Dynamic Path Loading. IEEE Transactions on Intelligent Transportation Systems 14 (2), 772-781. Celikoglu, H.B., 2013b. Reconstructing freeway travel times with a simplified network flow model alternating the adopted fundamental diagram. European Journal of Operational Research 228 (2), 457-466. Celikoglu, H.B., 2014. Dynamic Classification of Traffic Flow Patterns Simulated by a Switching Multimode Discrete Cell Transmission Model. IEEE Transactions on Intelligent Transportation Systems 15 (6), 2539-2550. Celikoglu, H.B., Silgu, M.A., 2016. Extension of Traffic Flow Pattern Dynamic Classification by a Macroscopic Model Using Multivariate Clustering. Transportation Science 50 (3), 966-981. FHWA, 2004. A Filed Test and Evaluation of Variable Speed Limits in Work Zones. http://safety.fhwa.dot.gov/speedmgt/vslimits/docs. Grumert, E., Ma, X. and Tapani, A., 2015. Analysis of a cooperative variable speed limit system using microscopic traffic simulation. Transportation research part C: emerging technologies, 52, pp.173-186. Hadiuzzaman, M., Fang, J., Karim, M.A., Luo, Y. and Qiu, T.Z., 2015. Modeling Driver Compliance to VSL and Quantifying Impacts of Compliance Levels and Control Strategy on Mobility and Safety. Journal of Transportation Engineering, 141(12), p.04015028. Hegyi, A., De Schutter, B. and Hellendoorn, H., 2005. Model predictive control for optimal coordination of ramp metering and variable speed limits.Transportation Research Part C: Emerging Technologies, 13(3), pp.185-209. Khondaker, B. and Kattan, L., 2015. Variable speed limit: A microscopic analysis in a connected vehicle environment. Transportation Research Part C: Emerging Technologies, 58, pp.146-159. Kotusevski, G. and Hawick, K.A., 2009. A review of traffic simulation software. Computer Science. Institute of Information and Mathematical Sciences, Massey University. Li, Z., Li, Y., Liu, P., Wang, W. and Xu, C., 2014. Development of a variable speed limit strategy to reduce secondary collision risks during inclement weathers. Accident Analysis & Prevention, 72, pp.134-145. Liu, B., Ghosal, D., Chuah, C.N. and Zhang, H.M., 2012. Reducing greenhouse effects via fuel consumption-aware variable speed limit (FCVSL). IEEE Transactions on Vehicular Technology, 61(1), pp.111-122. Liu, J., Wang, X.Y., Zhang, J.L. and Guo, D., 2012, August. Verification Method of Vehicle Driving Tendency Dynamic Recognition Model Based on Simulated Experiment under Car Following. In CICTP 2012@ sMultimodal Transportation Systems—Convenient, Safe, Cost-Effective, Efficient (pp. 2339-2349). ASCE. Long, S., Gentry, L. and Bham, G.H., 2012. Driver perceptions and sources of user dissatisfaction in the implementation of variable speed limit systems.Transport Policy, 23, pp.1-7. Nissan, A. and Koutsopoulosb, H.N., 2011. Evaluation of the impact of advisory variable speed limits on motorway capacity and level of service.Procedia-Social and Behavioral Sciences, 16, pp.100-109. Papageorgiou, M., Kosmatopoulos, E. and Papamichail, I., 2008. Effects of variable speed limits on motorway traffic flow. Transportation Research Record: Journal of the Transportation Research Board, (2047), pp.37-48. Ryus, P., Vandehey, M., Elefteriadou, L., Dowling, R.G. and Ostrom, B.K., 2011. New TRB Publication: Highway Capacity Manual 2010. TR News, (273). UK Highways Agency, 1996. UK Design Manual for Roads and Bridges. Vissim, P. T. V. "8 User Manual." PTV, Karlsruhe, Germany (2015). Yu, R. and Abdel-Aty, M., 2014. An optimal variable speed limits system to ameliorate traffic safety risk. Transportation research part C: emerging technologies, 46, pp.235-246.
