RESPONSE TIMES AND THE SAFETY OF EMERGENCY VEHICLES AT INTERSECTIONS

(1)

RESPONSE TIMES AND THE SAFETY OF EMERGENCY

VEHICLES AT INTERSECTIONS

A Simulation Study including Vehicles, Pedestrians, Bicyclists and Buses in the

Province of Groningen, the Netherlands

MASTER THESIS

MSc Technology and Operations Management

University of Groningen

Faculty of Economics and Business

January 2019

Linda Elisabeth Wittenhorst

S3271447

Praamstraat 40

8102 HN Raalte

l.e.wittenhorst@student.rug.nl

Company Supervisor Ir J. Hatenboer

(2)

MANAGERIAL INSIGHTS

The UMCG Ambulancezorg initiated this research project with the aim to yield insights into the effect of various types of road occupants; vehicles, buses, bicyclists and pedestrians, on the safety and the response time of an emergency vehicle (EV) at intersections. Previous research of the UMCG Ambulancezorg conducted by D. De Boer focused on the effect of the real-time traffic control system on the safety of emergency vehicles at intersections. In this project, the scope of the research was vehicles in the city of Groningen meaning that buses, bicyclists and pedestrians were excluded. As a follow up project, the UMCG Ambulancezorg requested a project in which the influence of these other road occupants is researched.

After completing a simulation study, comparing the EV detection distances of the types of road occupants in various combinations and determining the real-time control strategy in which all road occupants were included, the following can be concluded. First, the bus lane is the most successful lane for the emergency vehicle to use during the passage of the intersection. When the volume of the road occupants in increased with 250% and the EV speed is increase to the ultimate speed of 20 m/s the bus lane remains to be the most efficient lane. Due to the low bus volume and the separate lane for buses, the bus lane can be discharged smoother than the lanes for vehicles. Based upon this insight, this research advices to discuss the results with the local bus company and to make agreements in how EV’s can make use of bus lanes.

Second, the influence of pedestrians and bicyclists on the EV detection distance is significantly larger than that of buses. At an intersection in which pedestrians and bicyclists are included the EV detection distance needs to be 30% further in front of the intersection to ensure a safe ambulance passage. Thus, the advice of this research is to consider an intersection in which those road occupants are excluded. This could be in the form of a roundabout or tunnel.

(3)

simulate the EV passage under more realistic traffic situations. During this study, only one intersection is considered. Therefore, this research could be extended by evaluating the influence of multiple intersections or even the whole ambulance route. Last, this study only considered one priority request at the time. An extended study could be performed to research the real-time control strategy when more priority requests occur at one instance.

Summarizing, a real-time control strategy including all road occupants increases the safety of ambulances by providing, if necessary, green lights to the bus lane. This can decrease the response time of an ambulance, as it can cross the intersection at operating or even ultimate speed. Last, the influence of bicyclists and pedestrians on the EV detection distance is significant, when these road occupants are excluded from the intersection a safe ambulance passage and the response times will optimize.

(4)

ABSTRACT

(5)

ACKNOWLEDGEMENTS

Performing this research was not possible without the advice and support of a handful of people. First, I would like to express my gratitude to Dr E. Ursavas. With her directness and advice, she was able to guide me through this process. Also, I would like to thank Dr X. Zhu, my second supervisor, for his advice. Second, I would like to thank the UMCG Ambulancezorg for the opportunity to write my thesis for them. Especially, I would like to mention J. Hatenboer. He was able to provide useful insights for this research. Further, I would like to thank W. Ijdema from the Gemeente Groningen for gathering and providing me the data used for this research. Finally, I would like to thank my family and friends for their patience, support and motivational speeches during this period.

Raalte, January 2019

(6)

MANAGERIAL INSIGHTS ... 2 ABSTRACT ... 4 ACKNOWLEDGEMENTS ... 5 TABLE OF CONTENTS ... 6 1. INTRODUCTION... 7 2. LITERATURE REVIEW ... 9 3. METHODOLOGY ... 12 3.1 CASE DESCRIPTION ...12 3.2 SIMULATION MODEL...13 3.3 DATA COLLECTION ...14 4. SIMULATION ... 16

4.1 SIMULATION UNDER NORMAL CIRCUMSTANCES ...16

4.2 SIMULATION OF ROAD OCCUPANT VARIATION ...20

4.3 SIMULATION OF EV SPEED VARIATION ...21

4.4 SIMULATION OF VOLUME SATURATION ...22

5. DISCUSSION & CONCLUSION ... 24

REFERENCES ... 27

APPENDIX A – INPUT SIMULATION ... 29

(7)

1. INTRODUCTION

One of the crucial issues in the healthcare sector is the pre-hospital care provided by Emergency Medical Services (EMS). EMS provide first aid services in a timely and appropriate manner under emergency conditions (Aboueljinane, Sahin & Jemai, 2013). The response time of EMS is essential in improving the patients’ chances of survival and of meaningful recovery (Krishna, Kartha & Nair, 2017; Missikpode, Peek-Asa, Young & Hamann, 2018; Peleg & Pliskin, 2004).

When Emergency Vehicles (EV’s) are driving in emergency mode, traffic signals at intersections provide priority to the EV’s. However, when EV’s are driving in emergency mode and they use pre-emption at intersections, the risk of crashing and thereby injuring emergency vehicle occupants, occupants of other vehicles or other vulnerable road users increases (Missikpode et al., 2018). In the Netherlands, 34% of the accidents with EV’s occurred when an EV crossed an intersection (Kobes, Ros & Groenewegen-ter Morsche, 2017). Consequently, the response time of the EV’s increases and traffic congestion at the intersection occurs. Due to these consequences, it can be concluded that in order to achieve low response times for EV’s, a safe and efficient crossing of intersections is of high importance.

Previous research of Qin & Khan (2012) proposed two control strategies to optimize the response times of EV’s in a safe way and simultaneously minimizing the interference with the general traffic using traffic signal pre-emption. The first control strategy of Qin & Khan (2012) is the real-time control strategy, which provides green lights for the EV based on the real-time queue length and simultaneously reduce the interference with the other road occupants by allocating spare time to the side roads to minimize the queues. Alternatively, the fixed-time control strategy, administers a fixed time of green lights to the EV without providing extra green time to the side roads (Qin & Khan, 2012).

(8)

controls strategies (Wang, Ma & Yang, 2013; Mu, Liu & Li, 2018; Wang, Yun, Ma & Yang, 2013), none of these studies simulated a strategy including various types of road occupants. The Dutch traffic knows various types of road occupants; conventional vehicles, emergency vehicles, trucks, busses, trams, motorcyclists, bicyclists and pedestrians. All these road users behave differently, have various properties and traffic rules.

This paper builds upon the research of De Boer (2018) and therefore uses the real-time control strategy of Qin & Khan (2012). The aim is to yield insights on how different types of road occupants (vehicles, buses, bicyclists or pedestrians) influence a real-time traffic control strategy. In order to define how and to what extent different types of road users influence a real-time control strategy a central research question is phrased: “How and to what extent do

the vehicles, buses, bicyclists and pedestrians influence the safety and response times of a real-time control strategy when an Emergency Vehicle crosses an intersection?”

A simulation case study is used to analyse various situations in which different types of road occupants are present when an EV crosses an intersection. The UMCG Ambulancezorg Groningen provides the case study used for this research. Managerial relevance of this paper is gained by defining how and to what extent different types of road occupants influence a real-time control strategy to the UMCG Ambulancezorg Groningen. The academic relevance of this research is the extension of pedestrians, buses and bicyclists integrated into a simulation model to simulate a real-time control strategy. Also, the case study of the province of Groningen can be drawn wider as it represents the Dutch traffic.

(9)

2. LITERATURE REVIEW

This chapter provides an overview of relevant previous research. In academic research, various authors have proposed priority strategies for pre-emption of EV’s. Based upon these priority strategies numerous researchers proposed traffic control strategies for EV passage. These traffic control strategies or methods aim for low EV response times, a safe intersection passage and a minimal interference with other road users. However, none of these researchers simulated a traffic situation in which various road occupants were included.

In this research, priority is of importance to generate the pre-emption for the EV’s. Using simulation Wang, Ma and Yang (2013) proposed a degree-of-priority to classify priority vehicles and estimate the travel time which reduces the impacts on normal traffic by minimizing the queues. Also, Mu, Liu and Li (2018) developed an EV pre-emption control system, consisting of different subsystems; a monitoring subsystem, a phase time determination subsystem and a phase switching control subsystem. This research will extend the traffic priority strategies of Wang, Ma and Yang (2013) and Mu, Liu and Li (2018) not only providing priority to EV’s but also by including other road users.

(10)

Above real-time control model included conventional vehicles, but excluded other road users. However, it can be argued that other road users influence a traffic control strategy. Kockelman & Shabih (1999) studied that light-duty trucks generally need a significantly longer time to start the vehicle than a starting conventional vehicle. Due to the similar weight and length of light-duty trucks with busses, it can be reasoned that the starting time for buses is also significantly longer than that of a conventional vehicle. The research of Minh & Sano (2003) indicated that motorcycles strongly influence the traffic flow at signalized intersections when the motorcycle ration is high.

In the U.S. 40 percent of the collisions with pedestrians occur at intersections (Trentacoste, 2004). Hence, the safety of the pedestrians is of importance. Yang, Deng, Wang, Li & Wang (2009) simulated the road crossing behavior of pedestrians in a traffic system. This research was conducted in China and considered different states of pedestrians. The simulation of Yang et al. (2009) will be used to extend the real-time control strategy of Qin & Khan (2012) with pedestrians.

Ling & Wu (2004) studied the bicyclists’ behavior at signal-controlled intersections. Ling & Wu (2004) did not simulate their study; therefore, the observations about bicyclists’ behavior will be used to extent the real-time control strategy of Qin & Khan (2012). A conclusion of Ling and Wu (2004) is that the average speed of bicyclists is 3,06 m/s. Lovas (1994) researched that the average walking speed of pedestrians is between 1,30 m/s and 1,46 m/s, with the observation that the speed is influenced by personal and situational factors. Compared to the average speed of cars, which is 11,70 m/s, it can be argued that it takes more time for bicyclists and pedestrians to pass an intersection than conventional vehicles.

(11)

The real-time control strategy of Qin & Khan (2012) and De Boer (2018) will be used as design for this study. Including the above-mentioned strategies and observations about buses, bicyclists and pedestrians. Further, historical data is used as input for the simulation model to assess the influence of vehicles, buses, pedestrians and bicyclists on the real-time traffic control strategy. Given the input, the simulation model will determine the average queue lengths per hour. With this data the queue length at the moment the EV is detected can be calculated. Based upon this queue, the required notification time period and thereby the EV detection distance can be determined. The notification time period is the time necessary to clear the queue for the EV to pass the intersection. Consequently, the EV detection distance is the point in front of the intersection at which the EV should be detected in order to clear the queue in time for the EV to pass. Hence, the times required to switch the light to green, to discharge the queue and to include a safety time interval can be calculated. Finally, the green time required for the queue to discharge before the EV is approaching can be determined. The outcomes of this simulation model will provide insight in if and how the safety of an EV can be increased when all types of road occupants are present at the intersection. The conceptual model of this research is presented in Figure 2.1. The simulation model will be explained in more detail in chapter 3.

(12)

3. METHODOLOGY

To answer the research question posed in chapter 1, this chapter describes the methods used to gain insights in the real-time traffic control strategy. First, a description of the case is provided. Secondly, the simulation model is described. Finally, the manner in which data collection is presented in the third paragraph.

3.1 CASE DESCRIPTION

The UMCG Ambulancezorg provides the case for this research. UMCG Ambulancezorg is responsible for the ambulance dispatches in Drenthe and a part of Friesland. The UMCG Ambulancezorg aims low response times, in order to maximize the patient’s chances of survival. As mentioned before, the downside of a decrease in response time is the risk for collisions. In particular, the UMCG Ambulancezorg wants to focus on the influence of buses, pedestrians and bicyclists on a real-time traffic control strategy. Being the reason, the UMCG Ambulancezorg requested to gain insights into the effect of various types of road occupants; vehicles, buses, bicyclists and pedestrians, on the real-time traffic control system at intersections when an EV crosses. In the case of the UMCG Ambulancezorg the influence on the real-time traffic control system is measured by the influence of the safety of all road occupants and the efficiency in which the emergency vehicle can cross an intersection.

To yield the insights requested by the UMCG Ambulancezorg this research extends and adapts the model of Qin and Khan (2012) to analyse the impact of all road occupants in a real-time control strategy on the response time, the safety, efficiency and the interference with all other road occupants. Case data concerning the traffic and the characteristics of all road occupants is used. The model of Qin and Khan (2012) excludes buses, pedestrians and bicyclists; this research will include these road occupants in the simulation model. A simulation study suits this research as it enables to research various parameters and experiments without making investment and changes in the current traffic situation.

(13)

is the Griffeweg towards the Sontweg. Approach 3 is from the UMCG towards Europaplein. Finally, approach 4 refers to the Sontweg towards the Griffeweg. Additionally, approach 1 and 3 are considered main roads and approach 2 and 4 side roads.

FIGURE 3.1 – Intersection Europaweg – Griffeweg - Sontweg

Furthermore, the types of road occupants are abbreviated into V (vehicles), B (buses), C (bicyclists) and P (pedestrians). When there are more types of road occupants present in the scenario the abbreviations are combined. For example, a scenario with VBC includes vehicles, buses and cyclist. Furthermore, the vehicle lanes are named VL, VR, VM1 and VM2. VL indicates the vehicles on the left lane and VR represents the vehicles on the right lane. VM1 and VM2 are the two lanes in which the vehicles will go straight ahead. Please note, that approach 2 only exists of one lane, in which vehicles can go straight ahead hence, there is no VM2 present in approach 2.

3.2 SIMULATION MODEL

This research extends the research of Qin & Khan (2012) and De Boer (2018) by including buses, bicyclists and pedestrians in the intersection. To define how and to what extent the other road occupants influence the real-time control strategy for an EV to cross the intersection a simulation study will be performed in different phases.

(14)

percentages of vehicles turning left and right are used per approach. Secondly, the average queue can be calculated with the use of the road occupant arrivals. The average queue is determined for all types of road occupant per approach and in case of the vehicles per lane. Third, the EV detection distance is determined using the average queue length plus two times the standard deviation. The model calculated the distance from the intersection, where detection should be placed to ensure a safe passage for the EV. Finally, the real-time control strategy for the EV is simulated. This control strategy provides green lights for a period of time that is based upon the calculated time required to discharge the queue ahead of the ambulance, before it arrives at the intersection. After that, it calculates if there is spare time between the moment the queue is discharged and the EV arrives at the intersection. When the spare time exceeds a predefined standard, the real-time control strategy will allocate this spare time to the side road. This allocation of green light to the side road increases the efficiency of the intersection. For an explanation of the mathematics of the simulation model, please refer to Qin & Khan (2012).

3.3 DATA COLLECTION

The simulation model will use historical data about the intersection, provided by the municipality of Groningen. The data is collected from 8 October till 14 October 2018. October tends to be representative for all normal weeks as it is not a month known for people being on holiday (e.g. July, August, September and December). Additionally, the municipality checked for roadwork, that could influence the data but in the week of 8-14 October this was not the case. The data received from the municipality is processed before being used for the simulation.

First, the traffic signal timing plan is calculated for all road occupants according to Webster’s method as described in Qin & Khan (2012). The historical data of the intersection Europaweg-Griffeweg-Sontweg is used to determine the traffic signal timing plan. The results of the timing plan are depicted in Table 3.1.

TABLE 3.1 – Traffic Signal Timing Plan

Road Occupant Green Phase 1 Green Phase 2 Yellow Red

V 17 sec 14 sec 3 sec 1 sec

B 7 sec x 3 sec 1 sec

C 10 sec x 5 sec 1 sec

(15)

Green phase 1 refers to the time provided for the main road (approach 1 and 3) and green phase 2 indicates the time provided to the side roads (approach 2 and 4). The buses, bicyclists and pedestrians have no time registered for green phase 2, hence those road occupants receiving green time at the same moment.

Secondly, the volumes per road occupant and lane per hour per approach is calculated and depicted in Table 3.2. Unfortunately, the volumes for pedestrians are missing, due to limitations into the data gathering system of the municipality. Hence, these volumes are assumed on the reasoning that the volume of bicyclists is an indication for the volume of pedestrians.

TABLE 3.2 – Volumes (Road Occupants per Hour) per Lane per Approach

Approach VL VM1 VM2 VR B C P

1 94,99 250,98 250,98 168,47 15,68 48,19 30,00

2 132,30 137,35 x 86,28 7,37 37,74 25,00

3 139,62 256,01 256,01 170,62 11,35 37,77 25,00

4 119,06 71,41 71,41 136,72 7,37 56,32 40,00

Finally, the average speed per road occupant per approach is determined (Table 3.3). This is based upon the research of De Boer (2018) and Daamen (2003).

TABLE 3.3 – Average Speed (m/s) per Road Occupant per Approach

Approach V B C P

1 14,80 13,00 4,00 1,40

2 10,80 9,00 4,00 1,40

3 9,72 8,00 4,00 1,40

(16)

4. SIMULATION

The simulation model will be simulated using the software MATLAB (The Mathworks Inc., 2003). First, the simulation is performed under normal circumstances. In the normal situation, the historical data, the driving speed of the EV is 15 m/s, the EV is arriving from approach 1, the EV detection distance is 300 metres and all road occupants are included. Second, the variation of road occupants is simulated. Third, a simulation of variation in EV driving speed is carried out. Finally, the volume saturation is simulated. The simulations described in this chapter use parameters as characterized in Appendix A. An overview of all results is presented in appendix B. All simulations, will be simulated over 20 cycles. However, the real-time control strategy is not able to run over more than one cycle. Therefore, the results of these simulations are only over one simulation cycle. This is considered a limitation of this research.

The performance of the real-time control strategy is evaluated by the average queue, the EV detection distance, the green time and the success rate. All these parameters are performance indicators. The average queue indicates the number of road occupants that are still waiting in the lane. In this simulation, the lanes of vehicles and buses will be cleared, an EV cannot cross the intersection in the bicyclist- or pedestrian lane. The EV detection distance is the number of meters where the EV should be detected in front of the intersection to make sure the EV can cross the intersection in a safe manner. The green time is the time required to clear one of the vehicle- or bus lanes, before the EV is arriving. When the EV is detected on time to provide the required green time, the case will be considered successful. Contrary, when the EV cannot cross the intersection successfully, it will be called a failure. In this case, all lanes have a queue, there is no lane clear for the EV to cross the intersection. If this is the case, there is extra time needed to clear one queue.

4.1 SIMULATION UNDER NORMAL CIRCUMSTANCES

(17)

The results show that on average the buses and pedestrians have an average queue of 0 bus/pedestrian per hour for all approaches. Buses will show a queue when the volume for buses at approach 1 is 120 buses per hour. In the current situation, the volume is 15,68 buses per hour. Meaning, that the volume of buses must increase with 750% will there be an average queue of 0,1 bus per hour at approach 1. Cyclists, on the other hand, do have an average queue of 0,5 cyclists per hour for approach 1 and 4. This is caused by the larger volumes of cyclists for approach 1 and 4 with 48,19 and 56,32 cyclists per hour compared to the volumes of 37,74 and 37,77 for approach 2 and 3. The queues of vehicles show, as expected due to the higher volumes, that the main roads (approach 1 and 3) have a higher average queue than the side roads. Especially, the VM1 and VM2 of the main roads, the lanes for straight ahead, have a relatively high queue compared to the VL and VR. The side roads, on the other hand, have the largest queues for the lanes VL and VR. In this scenario, it can be argued that the bus lane is the best lane for the EV to use when crossing the intersection, as there is on average a queue of 0 busses per hour.

Figure 4.1 – Average Queue per Hour– Normal Circumstances

To simulate the EV detection distance the above calculated average queues are used. An overview of the results is provided in Table 4.1. In the current traffic situation, the intersection is receiving traffic light information 300 metres in advance. The results of the simulation are within the 300 metres; this means under normal circumstances the current traffic light software suffices. When the EV is arriving from one of the main roads the EV

0 0.5 1 1.5 2 2.5 3 VL VR VM1 VM2 B C P A ve rage Q u eu e (R o ad O cc u p an ts w ai tin g p er H o u r)

Type of Road Occupant / Lane

(18)

The entry moment at which the EV is detected can vary. The EV can be detected when the vehicles receive a green light at the main road or when they are at the middle of that green light, at the beginning of the yellow light or at the beginning of the red light. Moreover, the same scenarios can occur for the vehicles at the side roads, the buses, bicyclists and pedestrians. Meaning, there are 18 different scenarios at which traffic light moment the EV is detected; these are summarized in Table 4.2.

TABLE 4.2 - Emergency Vehicle Arrival Scenarios Scenario Entry Moment

1 At the beginning of green light Vehicles main road

2 Middle of green light Vehicles main road

3 The beginning of yellow light Vehicles main road

4 The beginning of red light Vehicles main road

5 At the beginning of green light Buses

6 Middle of green light Buses

7 The beginning of yellow light Buses

8 The beginning of red light Buses

9 At the beginning of green light Bicyclists

10 Middle of green light Bicyclists

11 The beginning of yellow light Bicyclists

12 The beginning of red light Bicyclists

13 At the beginning of green light Pedestrians

14 Middle of green light Pedestrians

15 The beginning of yellow light Pedestrians

16 The beginning of red light Pedestrians

17 At the beginning of green light Vehicles side road

18 Middle of green light Vehicles side road

For all 18 traffic light scenarios, at which the EV can be detected, the real-time control strategy is simulated under normal circumstances. Meaning, that he volumes of road users are based upon the historical data, the driving speed of the EV is 15 m/s, the EV is arriving from

TABLE 4.1 – EV Detection Distance (Metres) – Normal Circumstances

1 2 3 4

(19)

approach 1, the EV detection distance is 300 metres and all road occupants are included. The results of the simulation, presented in Table 4.3, show that traffic light scenarios 1 and 4 till 18 have the same results. In these scenarios, the real-time control strategy cannot discharge the VM1 and VM2 lane on time for the EV to pass. For these lanes, the EV detection distance has to be enlarged in order to discharge the lanes on time. The VL and bus lane, on the other hand, can discharge their queue on time. Hence, these lanes are considered successful. After, the VL lane has discharge its queue, this lane has a spare time of 5,60 seconds left. The bus lane does not need green time for the main road, therefore, this time of 11,00 seconds is provided to the side roads. Contrary, the VR lane fails to discharge the queue on time, the queue indicates that there is still one vehicle in queue. In order to discharge this queue there is 0,60 seconds extra time needed. For traffic light scenario 2 and 3, at which the vehicles are receiving a green light when the EV is detected, the results are different from the other scenarios. Still, for VM1 and VM2 is the EV detection distance of 300 not enough. In these scenarios, the VR lane is still failing, however, the extra time needed has increased to 1,90 seconds. Finally, the VL and bus lane can successfully discharge their queue. The VL lane has a spare time of 5,00 seconds and the bus lane can provide 11 seconds to the side roads.

TABLE 4.3 - Real-time Control Strategy – Normal Circumstances

Traffic Light

Scenario Lane Successful

Green Time Main Road (Seconds) Green Time Side Road (Seconds)

Time Needed/ Spare

Time (Seconds) Queue

1, 4-18

VL Yes 11,00 0,00 5,60 0,00

VR No 11,00 0,00 0,60 1,00

VM1 No, because EV Detection Distance is not long enough

B Yes 0,00 11,00 0,00 0,00

2, 3

VL Yes 11,00 0,00 5,00 0,00

VR No 11,00 0,00 1,90 1,00

B Yes! 0,00 11,00 0,00 0,00

(20)

minimize the interference with other road occupants and delay for other lanes. Further, for the EV to use the VM1 or VM2 lane for intersection passage the EV detection distance has to be enlarged. Finally, when the EV is detected when the traffic lights have started to be green for vehicles this will result into a lower spare time for VL and an increase in extra time needed for VR.

4.2 SIMULATION OF ROAD OCCUPANT VARIATION

In this paragraph, the EV detection distance is simulated for all various road occupant combinations. The normal circumstances remain the same, only the types of road occupants included in the intersection varies. In the simulation vehicles will always be present, as the aim of this study is to determine the influence of busses, bicyclists and pedestrians on the intersection passage of the EV. Therefore, the following road occupants variations are simulated; VBCP, VBC, VBP, VCP, VB, VC, VP. The EV detection distances when the EV is arriving from approach 1 for all variations is displayed in Table 4.4.

TABLE 4.4 – EV Detection Distance (Metres) – Road Occupant Variation

VBCP VBC VBP VCP VB VC VP V

236,9 223,4 229,3 231,8 158,8 231,7 229,3 148,5

(21)

4.3 SIMULATION OF EV SPEED VARIATION

In this paragraph, the influence of a change in EV speed on the EV detection distance is simulated. According to the Instituut Fysieke Veiligheid the speed limit of an EV’s is 20 m/s (Kobes, Ros & Groenewegen-ter Morsche, 2017). The average speed of an EV, on the other hand, is 13 m/s. Therefore, in this simulation; it will be tested if an EV can cross the intersection at an operating speed of 15 m/s. Furthermore, the ultimate EV speed limit of 20 m/s will also be tested in this simulation, this because UMCG Ambulancezorg aims to reach the hospital as soon as possible. Hence, the variation in EV speed is simulated for the average speed of 15 m/s and the ultimate speed of 20 m/s, all other normal circumstances remain the same. The results of this simulation are provided in Table 4.5.

TABLE 4.5 – EV Detection Distance and Time per Approach– EV Speed Variation

1 2 3 4

EV Detection Distance (Metres)

15 m/s 236,89 192,03 228,20 165,00

20 m/s 315,86 256,04 304,30 220,00

EV Detection Time (Seconds)

15 m/s 15,79 12,80 15,20 11,00

20 m/s 15,79 12,80 15,20 11,00

As expected, the EV detection distances are lower when the EV is driving 15 m/s than when the driving speed of the EV is 20 m/s. When the EV is driving faster more metres are covered within the same time. As the simulation results show, the EV detection time does not change when the EV driving speed increases. Consequently, to cover the increase in speed, the EV detection distance must be longer. The increase in speed of 5 m/s (33%) results in an increase in metres for EV detection distance of 33% for all EV arrival approaches.

(22)

To conclude, an increase in EV driving speed to 20 m/s does have an influence on the real-time control strategy, however the VL and bus lane can still be cleared successfully.

TABLE 4.6 - Real-time Control Strategy – EV Speed Variation

EV Driving

Speed Lane Successful

15 m/s

VL Yes 11,00 0,00 5,60 0,00

VR No 11,00 0,00 0,60 1,00

B Yes 0,00 11,00 0,00 0,00

20 m/s

VL Yes 6,00 0,00 0,40 0,00

VR No 6,00 0,00 6,90 1,00

B Yes! 6,00 0,00 6,00 0,00

4.4 SIMULATION OF VOLUME SATURATION

Following, the influence of traffic volume saturation on the EV detection distance is simulated. In all above simulations the historical volumes received from the municipality were used. In this paragraph, the historical volumes are compared to a decrease of 50% and an increase of 250% (Table 4.7). All other normal circumstances remain the same.

TABLE 4.7 – EV Detection Distance (Metres) per Approach – Volume Saturation Volume Saturation 1 2 3 4 Historical Volumes (100%) 236,89 192,03 228,20 165,00 50% 184,50 (-) 22% 165,00 (-) 14% 192,60 (-) 16% 165,00 0% 250% 537,70 (+) 127% 215,80 (+) 13% 413,90 (+) 81% 209,90 (+) 27%

(23)

This is as expected due to the fact that also the detection distance is the biggest of all for EV arrival approaches. The EV detection distance differences in volume saturation for the side roads is relatively small.

Finally, the volumes are saturated for the real-time control strategy. The results in Table 4.8 show, that when the volume is decreased with 50%, all lanes can be discharged on time. Even, all lanes have a spare time. Varying from 11 seconds for VL and the B lanes, to 7,40 seconds for VR and 5,60 seconds for the lanes VM1 and VM2. The spare time of the bus lane is allocated to the side roads. When, the volume increased the VR, VM1 and VM2 lanes have a 100% failure. For these lanes, the EV detection distance is not long enough. The VL lane can discharge the queue on time, however the spare time is decreased to 7,30 seconds. Finally, the bus lane can discharge its queue and even has the same spare time compared to the 50% situation. Based upon this it can be argued that the bus lane is the best lane for the EV to use in an intersection passage, even when the volume is saturated.

TABLE 4.8 - Real-time Control Strategy – Road Occupant Volume Saturation

Volume

Saturation Lane Successful

Historical Volumes

(100%)

VL Yes 11,00 0,00 5,60 0,00

VR No 11,00 0,00 0,60 1,00

B Yes 0,00 11,00 0,00 0,00 50% VL Yes 11,00 0,00 11,00 0,00 VR Yes 11,00 0,00 7,40 0,00 VM1 Yes 11,00 0,00 5,60 0,00 VM2 Yes 11,00 0,00 5,60 0,00 B Yes 0,00 11,00 0,00 0,00 250% VL No 11,00 0,00 7,30 1,00

VR No, because EV Detection Distance is not long enough

(24)

5. DISCUSSION & CONCLUSION

After performing all simulations in the various scenarios, the results will be discussed and conclusions will be drawn. First, under normal operation circumstances, the average queue of buses is zero per hour for all approaches. Hence, the bus lane would be the safest lane for the EV to use for intersection passage. This result is supported by the results of the real-time control strategy under normal circumstances, that simulated that the bus lane can be successfully cleared in all traffic light scenarios. Moreover, the real-time control strategy also simulated that the bus lane can be cleared successfully when the EV driving speed has increased to 20 m/s and the volume of road occupants is 250%. Based upon these insights, it can be concluded that under all simulations the bus lane is a successful lane for the EV to use for intersection passage.

The vehicle lanes, VM1 and VM2 tend to result in failures in all scenarios. The simulation indicates that the EV detection distance of 300 metres is not long enough for the lane to be discharged before the EV arrives. The VR lane can be discharged before the EV arrives when the volume is 50%. However, under normal circumstances, when the volumes increase or when the EV driving speed increases, the VR cannot be discharged on time. The extra time needed, for the VR lane to discharge its queue before the EV arrives varies from 0,60 seconds under normal circumstances to 7,30 seconds when the volumes has increased to 250%. Contrary, the VL lane can discharge its lane successfully before the EV arrives under normal circumstances, when the volume is 50% and when the EV speed is 20 m/s. Nevertheless, when the volume increases to 250% the VL lane needs 7,30 second extra to be able to discharge its lane before the EV arrives. To conclude, the bus lane is the most successful lane for the EV to use for intersection passage. As, the average queue per hour for buses lower than the queue of the vehicle lanes.

(25)

Secondly, the results of the EV detection distances in which all road occupants are included shows bicyclists and pedestrians do have an influence on the EV detection distance. As a matter of fact, the EV detection distances of VC and VP are 30% longer than that of the scenario including VB. Also, the EV detection distances of VBC and VBP, in which a B are present is lower than that of VCP. Hence, the impact of buses on the intersection situation is lower than that of bicyclists and pedestrians. Based upon the EV detection distance it can be concluded that the impact of bicyclists and pedestrians is equal. This could be explained due to the fact that volume of bicyclists per hour is higher than that of pedestrians but pedestrians have a lower walking speed than bicyclists.

Theoretical implication of this research is the applicability of the real-time control strategy in which vehicles, buses, bicyclists and pedestrians are included for the traffic situation in the Netherlands. This study has addressed the influence and applicability of buses, bicyclist and pedestrians on the real-time control strategy. Hence, this paper adds to the existing field of research by providing insight into the influence of all road occupants on the safety and response time of EV’s. Also, this paper provides initial insight into a real-time traffic control strategy.

(26)

Although the above-described insights, there are some limitations to this research. The real-time control strategy of this study was not able to run the simulation over more runs. To gather a more reliable and valid results the real-time control strategy simulation should be extended to more runs. Further, the study uses historical data gathered over one week. When this is collected as a more extensive (i.e. longer) dataset this could provide more optimal parameters to simulate the EV passage under more realistic traffic situations. During this study, only one intersection is considered. Therefore, this research could be extended by evaluating the influence of multiple intersections or even the whole ambulance route. Last, this study only considered one priority request at the time. An extended study could be performed to research the real-time control strategy when more priority requests occur at one instance.

This study started in chapter 1 with the following research question: “How and to what extent

do the vehicles, buses, bicyclists and pedestrians influence the safety and response times of a real-time control strategy when an Emergency Vehicle crosses an intersection?” After

(27)

REFERENCES

Aboueljinane, L., Sahin, E. and Jemai, Z., 2013. A Review on Simulation Models applied to Emergency Medical Service Operations. Computer & Industrial Engineering, 66, 734-750.

Daamen, W. and Hoogendoorn, S.P., 2003. Experimental Research of Pedestrian Walking Behavior. Transportation Research Board Annual Meeting. 2003, 1-6.

De Boer, D., 2018. Increasing Safety through Technology implementation at Intersections: A Simulation Study with Emergency Medical Services in the Northern part of the Netherlands.

Kobes, M., Ros, A. and Groenewegen-ter Morsche, K., 2017. Ongevallenstatistiek Voorrangsvoertuigen 2014-2015. Instituut Fysieke Veiligheid.

Kockelman, K., Shabih, R.A., 1999. Effect of Vehicle Type on the Capacity of Signalized Intersections: The Case of Light-Duty Trucks. Journal of Transportation Engineering, 126 (6).

Krishna, A., Kartha, B.A. and Nair, V.S., 2017. Dynamic Traffic Light System for Unhindered Passing of High Priority Vehicles. Global Humanitarian Technology

Conference.

Ling, H. and Wu. J., 2004. A Study on Cyclist Behavior at Signalized Intersections. IEEE

Transactions on Intelligent Transportation Systems, 5(4), 293-299.

Lovas, G.G., 1994. Modeling and Simulation of Pedestrian Traffic Flow. Transportation

Research Part B: Methodological, 28 (6), 429-443.

(28)

Minh, C.C., and Sano, K., 2003. Analysis of Motorcycle Effects to Saturation Flow Rate at Signalized Intersection in Developing Countries. Journal of the Eastern Asia Society for

Transportation, 5, 1211-1222.

Missikpode, C., Peek-Asa, C., Young, T. and Hamann, C., 2018. Does crash risk increase when emergency vehicles are driving with lights and sirens? Accident Analysis and

Prevention, 113, 257-262.

Mu, H., Liu, L. and Li, H., 2018. Signal Pre-emption Control of Emergency Vehicles Based on times Coloured Petri Nets. Discrete Dynamics in Nature and Society, 2018.

Peleg, K. and Pliskin, J.S., 2004. A Geographic Information System Simulation Model of EMS: Reducing Ambulance Response Time. The American Journal of Emergency

Medicine, 22(3), 165-170.

Qin, X. and Khan, A.M., 2012. Control Strategies of Traffic Signal Timing Transition for Emergency Vehicle Pre-emption. Transportation Research Part C, 25, 1-17.

Trentacoste, M., 2004. Pedestrian and Bicyclist Intersection Safety Indices. US. Department

of Transportation Federal Highway Administration.

Yang, J., Deng. W., Wang, J., Li, Q. and Wang, Z., 2006. Modelling pedestrians’ road crossing behavior in traffic system micro-simulation in China. Transportation Research Part

A, 40, 280-290.

Wang, J., Ma, W. and Yang, X., 2013. Development of Degree-of-Priority Based Control Strategy for Emergency Vehicle Pre-emption Operation. Discrete Dynamics in Nature and

Society.

(29)

APPENDIX A – INPUT SIMULATION

Table A.1. Parameters for Simulation

Parameter Value

Saturation 2000 Road occupants/hour

Lost time per phase 3 Seconds

All red time

- Vehicles and Buses

- Bicyclists and Pedestrians

3 Seconds 5 Seconds

Lost time per cycle 8 Seconds

(30)

APPENDIX B – RESULTS NORMAL CIRCUMSTANCES

Table B.1 – Average Queue Length under Normal Circumstances Approach 1

(31)

(32)

(33)