INCREASING SAFETY THROUGH TECHNOLOGY
IMPLEMENTATION AT INTERSECTIONS
A Simulation Study with Emergency Medical Services in the Northern part of the
Netherlands
Master Thesis, MSc Supply Chain Management University of Groningen, Faculty of Economics and Business
19 June 2018 Dorien de Boer Mutua Fidesstraat 17 9741CB, Groningen S2473704 d.de.boer.15@student.rug.nl
Supervisor from the University of Groningen Dr E. Ursavas
Co-assessor Dr I. Bakir
3
MANAGERIAL INSIGHTS
This project was initiated by the UMCG Ambulancezorg with the aim to gain insight into the effect of real-time traffic control systems on the safety of emergency vehicles at intersections. After a simulation study is performed comparing the fixed time control strategy to a soon to be implemented real-time control strategy, the following can be concluded. Using a real-time control strategy an emergency vehicle can cross an intersection safely as a real-time control system provides a green light to an emergency vehicle. However, this has some restrictions. First, the emergency vehicle detection distance should be larger than 300 metres. With this distance less than 300 metres, the real-time control strategy occasionally fails to provide a green light. Further, this study is solely performed for ambulances with an operating speed of 15 and 20 m/s.
Supplementary to an increase in safety, the real-time control strategy minimises the total delay of other vehicles at the intersection. This is supported when the emergency vehicle detection distance is between 300 and 500 metres. If the real-time control strategy has an emergency vehicle detection distance of more than 500 metres, the average total delay starts to increase. The above-described results are confirmed for historical data, but also for situations with a higher traffic density (until 1500 vehicles per hour). When the arrival of vehicles should exceed 2000 vehicles per hour, the total delay starts to increase tremendously. Considering the characteristics of the analysed intersection and the city, it is not likely this density will occur.
As described above, the software implemented in the intelligent traffic light control should be able to detect and start to adjust its process before the ambulance drives into a range of 300 metres of the traffic light. When software is not able to start this adjustment process so far ahead, a software extension is required. Added, when the emergency vehicle detection distance exceeds 500 metres, the average delay starts to increase. This does not result in optimal efficiency and as can be noted. Therefore, considering changing traffic situations, it is not recommended to set an emergency vehicle detection distance larger than 500 metres.
4
optimal period of estimated time of arrival could be set lower. In addition, with an operating speed of 15 m/s and an emergency vehicle detection distance of 900 metres, the estimated time of arrival is 60 seconds. It can be noted that in a period of 60 seconds the traffic situations at the intersection are likely to have changed.
Although the above-described insights are provided, there are some limitations to this paper. During the evaluation of both strategies, only emergency vehicles and conventional vehicles are considered. Therefore, the evaluation could be improved by including buses, bicycles, and pedestrians into the simulation. A more accurate emergency vehicle detection distance could be provided. Further, an extended study could be performed to study the most efficient allocation of green light when multiple priority requests occur at one instance. Last, an optimised traffic flow could be improved by looking at the route of an ambulance instead of one intersection. This could assure an ambulance to reach its destination without decreasing the speed while minimizing total traffic delay at multiple intersections.
5 ABSTRACT
6
ACKNOWLEDGMENTS
This thesis reflects my final part of the master Supply Chain Management at the University of Groningen. This paper is the result of hard work and persuasion of the past few months. I would like to express my gratitude towards the people who have helped me to finish this project. First, I would like to thank my supervisor Dr E. Ursavas for her advice during the thesis project. Second, I am grateful for the opportunity to write my thesis for the UMCG Ambulancezorg under the supervision of Ir J. Hatenboer and H. van Ommen. They have helped me to gain insights into the industry of Emergency Care. Further, I would like to thank Gemeente Groningen, especially W. IJedma, for the cooperation and providing me with data for the simulation performed. Added, I especially would like to show my gratitude towards Professor X. Qin and Professor A.M. Khan for the approval to work with their developed model. Last, I would like to thank my boyfriend, family and friends for their support during this period.
Groningen, June 2018
7 TABLE OF CONTENTS Managerial Insights ... 3 Abstract ... 5 Acknowledgments ... 6 1. Introduction ... 9 2. Literature Review ... 10
2.1 Emergency Medical Services ... 10
2.1.1 Sector ... 10
2.1.2 Process ... 11
2.1.3 Performance indicators ... 11
2.2 Development of Technologies in Traffic Control ... 13
2.2.1 Traffic light ... 13
2.2.2 Connected traffic control ... 14
2.2.3 Intelligent control systems ... 15
3. Description of Control Strategies ... 18
3.1 Research Setting ... 18
3.2 Simulation Model ... 19
3.2.1 Generation of vehicles ... 19
3.2.2 Emergency vehicle distance ... 20
3.2.3 Control strategies ... 20
3.2.4 Intersection layout ... 22
3.3 Data Collection ... 22
4. Simulation ... 24
4.1 Analysis using Historical Data ... 25
4.1.1 Queue length under normal operation ... 25
4.1.2 Simulation of fixed time control ... 26
8
4.2 Analysis with Highly Dense Traffic ... 28
4.3 Emergency Vehicle arriving from Side Road ... 29
5. Discussion ... 30
5.1 Comparison Control Strategies EV from Main Road ... 30
5.2 Comparison Control Strategies EV from Side Road ... 32
5.3 Value of Real-Time Control Strategies ... 33
6. Conclusions ... 35
References ... 37
Appendix A Input Simulation ... 41
Appendix B Results Emergency Vehicle from Main Road ... 42
Appendix C Results Highly Dense Traffic ... 46
Appendix D Results with Emergency Vehicle from Side Road ... 48
9
1. INTRODUCTION
Accidents with priority vehicles occur mostly in surroundings where people feel safe, e.g. in their own neighbourhood or at intersections (Burke, Salas & Kincaid, 2001; Sanddal, Sanddal, Ward & Stanley, 2010; Instituut Fysieke Veiligheid (IFV), 2017). Most of these accidents (34%) occur when a priority vehicle passes a red light while the other vehicle receives a green light (Instituut Fysieke Veiligheid, 2017). Priority vehicles need to move to an emergency in a safe and timely matter, however, accidents delay this process. Quick response to the patient is considered important for emergency vehicles and therefore response time is considered as a key performance indicator for the EMS sector (Lam et al., 2015; Fassbender, Balucani, Walter, Levine, Haass & Grotta, 2015; Missikpode, Peek-Asa, Young & Hamman, 2018).
Previous papers have studied several ways to reduce response times. Peleg and Pliskin (2004) researched the response times using Geographic Information Systems and compared the response times in rural and urban areas. Papers have been dedicated to study the allocation and relocation of ambulances to reduce response times (Aboueljinane, Sahin, Jemai & Marty, 2014; Lam et al., 2015; McCormack, Coates, 2015; Zaffar, Rajagopalan, Saydam, Mayorga & Sharer, 2016). Further, Pinto, Silva & Young (2015) proposed a generic simulation model to improve the performance of ambulances. Linganagouda, Raju & Patil (2016) and Nellore and Hancke (2016) studied the consequences of traffic congestion on response time.
Although numerous studies have researched the reduction of the response time of emergency vehicles, studies are not abundantly conducted on increasing the safety of emergency medical services. Whereas Krishna, Kratha & Nair (2017) and Jin, Ma & Kosonen (2017) studied the dynamic allocation of traffic lights for conventional vehicles given connected vehicle technologies, this dynamic allocation has not been studied including priority requests of emergency vehicles.
10
studied if and under what circumstances an emergency vehicle can cross the intersection at operating speed. The second goal is to evaluate if the emergency vehicle can cross the intersection while minimising the interruption of the conventional traffic flow. Therefore, this paper addresses the following research question:
What is the influence of a real-time control strategy on the safety and smoothness of Emergency Medical Services crossing an intersection?
The academic relevance is provided by adapting the real-time control strategy to Dutch traffic characteristics, while updating the model considering recent technology improvements. Performing a simulation study, it is analysed if the real-time control strategy ensures a safer operation for ambulances while minimizing traffic delay. The relevance for management is found in the evaluated applicability of a real-time control strategy with the aim to increase the safety of emergency medical services. It is tested, given the software developments of traffic light control, if an emergency vehicle can increase the safety of its operation by influencing the traffic lights at an intersection.
The rest of this paper is organized as follows. In section two a literature review on the Dutch emergency medical service sector and traffic light developments is provided. Section three presents the simulation model and the improvements, where section four presents the results. Section five elaborates on the implications of the results and, last, section six presents the conclusions of this research and indicates further research opportunities.
2. LITERATURE REVIEW
2.1 Emergency Medical Services
2.1.1 Sector
11
over 1.3 million times in 2016 (Ambulancezorg, 2017; Ambulancezorg Nederland, 2018). From these dispatches, 632.875 are dispatched under category A1, the most emergent dispatch (Aringhieri et al., 2013; Ambulancezorg, 2013; Reuter-Oppermann, van den Berg & Vile, 2017). Ambulances driving under A1 statues are allowed to use sirens and lights and have right of way. Apart from the responsibilities, the EMS in the Netherlands is similar to other countries in Europe (Bos, Krol, Veenvliet & Plass, 2015).
2.1.2 Process
According to Aboueljinane, Sahin & Jemai (2013) three characteristics can be considered as important when modelling the demand of EMS sector, namely arrival distribution, geographical distribution and priority of calls. An ambulance under emergency has no scheduled distribution throughout the day as for instance buses. Whereas the arrival distribution is considered the moment an ambulance arrives at a specific location or intersection, this process of the emergency vehicle starts at the dispatch.
The process of an ambulance dispatch starts when a patient calls the emergency number. The centralist answers the call at one of the 19 central operations centres and assesses the severity of the situation, the triage (Aringhieri et al., 2013; Ambulancezorg, 2013; Reuter-Oppermann, et al., 2017). Based on the triage the ambulance dispatch is either placed in the A1 or A2 category, the priority of the call (Aboueljinane et al., 2013). A2 dispatches are those where a patient requires treatment, although the situation is not life-threatening. However, this study merely focuses at A1 dispatches as Missikpode et al. (2018) identified most of the fatal accidents involve A1 ambulances (46%). After the need of an ambulance is confirmed and triage is finished, the centralist assigns an ambulance to the emergency from a specific location (e.g. geographical distribution) (Aringhieri et al., 2013; Ambulancezorg, 2013). From the operation centre relevant information is transferred to the assigned ambulance and employees, and the driver goes to the emergency. At the location the employees of the ambulance will provide the initial treatment. If needed the patient will be transferred from the emergency location to another care provider (e.g. hospital). When the patient is either transferred to another care provider or does not need additional care, the process of the ambulance care ends (Ambulancezorg, 2013). An overview of the process is provided in figure 2.1.
2.1.3 Performance indicators
12
(e.g. response time). The quality of the process ultimately has an influence on the outcome quality (e.g. patient survival). Patient survival is one of the main goals of emergency medical services, however difficult to use as a performance measure since this is only assessed when the patient leaves the medical care system (McLay & Mayorga, 2010; Aboueljinane et al., 2013; Zaffar et al., 2016). As the EMS has merely influence on the first couple of minutes, patient survival does not provide an accurate reflection of the performance.
Fig. 2.1. Ambulance dispatch process (adapted from Ambulancezorg Nederland, 2018).
Whereas patient survival is difficult to use, response time is a key performance indicator (Aringhieri et al., 2013; Stein, Wallis & Adetunji, 2015; Zaffar et al., 2016). This because a prompt response to injuries influences strongly the survival or efficient recovery of a patient and EMS have an influence on this response period (Aringhieri et al., 2013). Response time covers the period between the moment an emergency call is answered until the ambulance is at the emergency scene. This is indicated in figure 2.1 as the grey arrow (Lam et al., 2015; Reuter-Oppermann et al., 2017). The threshold to measure performance of EMS has been defined as a response time of 8 minutes and 95% of the calls must be reached in 10 minutes, whereas in the Netherlands 95% of the calls require to have a response time less than 15 minutes (Eisenberg et al., 1979; Pons et al., 2005; Ambulancezorg Nederland, 2018). The overall threshold for response time varies in Europe between 5 and 20 minutes (Bos et al., 2015). Lam et al. (2015) identified several factors having a negative influence on the response time of ambulances. One of these factors is ‘process characteristics’ including place of incident, weather, and traffic (Lam et al., 2015).
Patient calls 112 Ambulance process ends Operation Centre answers call Triage Assign Ambulance Ambulance driving to scene Arriving at emergency scene Providing treatment at location Transferring patient to care provider Patient does not require extra care
13
For the scope of this study, the focus is on the risk factor traffic as increasing density of traffic has a significant impact on the response time of ambulances (Nellore & Hancke, 2016). Further, accidents occurring in traffic delay the process of an emergency medical service significantly.
Supplementary to the above-described characteristics, Ambulance Traffic Accidents (ATAs) also influence the response time negatively. ATAs prevent ambulances from being at the emergency location as fast as possible. Compared to accidents with solely passenger cars, ATAs are considered more severe as these result 1.7 times more often in a fatal accident (Chiu et al., 2018). Several factors can be identified that increase the occurrence of an accident. Missikpode et al. (2018) identified one of these factors as ‘failure to follow
traffic signals’. As in line with this factor, most accidents with emergency vehicles (EV) at
intersections occur when the EV crosses a red light while other vehicles cross a green light (82%) (IFV, 2017; Slattery & Silver, 2009). Although EV have the right to cross the red light when driving under emergency status, it is the main cause of accidents as passenger cars might have missed the approaching ambulance. Further, Missikpode et al. (2018) identify the risk factor ‘speeding’. The maximum speed crossing a red light with an emergency vehicle is 20 km/h (e.g. 5.50 m/s). However, most accidents where speed restrictions were neglected occur at intersections where the EV crosses a red light (95%). The average speed was 26 km/h (e.g. 7.22 m/s) higher than accepted (Instituut Fysieke Veiligheid, 2017). For a complete overview of risk factors, please refer to Missikpode et al. (2018).
2.2 Development of Technologies in Traffic Control
2.2.1 Traffic light
14
(2012) non-pavement invasive technologies can consist of microwave radar, active and passive infrared-based equipment, ultrasonic and acoustic detectors. However, these non-pavement invasive technologies are not always accurate due to missed priority requests and sensitivity to severe weather conditions (Yang et al., 2016).
The limitation of both technologies is the fixed sensor near intersections, and therefore, are not able to include real-time data from vehicles further away from the intersection. Currently, most intersections in the Netherlands make use of loop detectors in the road for priority vehicles (IJedema, 2018). However, these systems usually handle one signal at the time. When multiple vehicles approaching an intersection and sending priority signals, the traffic light will handle the first signal received first. Therefore, current signal pre-emption technologies result in suboptimal results (He et al., 2014). Recent technologies allow testing new methods for more efficient traffic control.
2.2.2 Connected traffic control
15
Fig. 2.2. iVRI technology (adapted from Talking Traffic, 2018).
An iVRI consists of the following. First, an iVRI consist of a Traffic Light Controller (TLC) which oversees and controls the traffic lights. A Road Site Information Technology Service station (RIS host) is installed as a basis to run the Information Technology Service (ITS) application (Talking Traffic, 2018). The ITS application informs, optimises and prioritises the traffic light based on received data. In addition, the iVRI can be connected with Wi-Fi which enables the iVRI to directly communicate with vehicles or people near the intersection (Talking Traffic, 2018). An important aspect of the implementation of iVRI is the prioritising of vehicles at intersections. Using data from connected vehicles, mobile phones and GPS systems, the traffic light maintains real-time data about all approaching road users. Noticing in advance what priority vehicles are approaching, a more efficient green light allocation is facilitated to corresponding approaches. This could help to increase the efficiency of an intersection while decreases the queue lengths. Another aspect is that cities can allocate different prioritising rules to intersections. Intersections with a high density of approaching trucks, could be instructed to more priority to trucks to reduce CO2 emission (He et al., 2014). Although this technology appears to be promising, the limitation of the software implemented is that for the initial phase, the traffic light cannot look more than 300 metres ahead (Meulenaere, 2018). This might impair the optimization of smooth traffic control.
2.2.3 Intelligent control systems
16
rate of connected vehicles will not be reached in the coming years (IJedema, 2018). Feng, Head, Khoshmagaham & Zamanipour (2015) designed a real-time adaptive signal phase allocation algorithm assuming connected vehicle data is available. Similar to other algorithms in the field, the objective is to minimize total vehicle delay and queue length at an intersection. This algorithm considers various penetration rates of connected vehicles. Later work of Yang et al. (2016) provide an improved model of Guler et al. (2014) assuming a certain percentage of vehicles is automated. The paper introduces an algorithm for intersections that recognizes three different types of vehicles, e.g. conventional, connected and automated. The model demonstrates that the decrease in delay diminishes after a certain penetration rate of connected vehicles. For a complete description of the algorithm, please refer to Yang et al. (2016). The common limitation of the above-described models is that all models ignore priority requests.
17
The overall limitation of above-described studies is the use of estimated data. As stated by Aringhieri, Bruni, Khodaparasti, van Essen (2017) recent technologies and connected cities provide opportunities to develop more realistic models based on real-time data. Models provide optimal results based on a certain volume of traffic, however, this demand is fluctuating during various periods of a day (Jin et al., 2017). Several studies have been dedicated to creating algorithms and simulations to improve the performance of emergency medical services. However, as stated by Aringhieri et al. (2017) these allocations, dispatching and routing models, are all aimed to improve performance regarding the patient survival.
The aim of this paper is, while adapting a real-time control strategy, to gain insight into the safety and smoothness of Emergency Medical Services crossing an intersection. It is studied if and under what circumstances an emergency vehicle can cross the intersection at operating speed. The second goal is to evaluate if the emergency vehicle can cross the intersection while minimising the interruption of the conventional traffic flow. This paper contributes to the field by providing insights on the applicability of real-time control strategies in the Netherlands. Building upon the model of Qin & Khan (2012), a case study is performed using the traffic characteristics of the city of Groningen. This paper builds upon the model of Qin and Khan (2012) as they propose a real-time control strategy to provide priority to the emergency vehicle with the aim of increasing its safety. Using Dutch traffic characteristics the model it is tested if the model can be applied in the Netherlands. In addition, given technology limitations, simulations are performed if the real-time strategy can be applied in the Netherlands. The real-time control strategy is compared to a fixed time control strategy (i.e. the base case).
18
These characteristics influencing the operation of an emergency medical service are used as input for the simulation study performed. Given this input, the simulation model will calculate if an approaching emergency vehicle can cross the intersection or if more time is required. Furthermore, under a real-time strategy the simulation model will calculate if other approaches can receive green light to minimize the total traffic delay. The outcomes of these simulations will provide insight if and how the safety of an emergency medical service can be increased. The conceptual model is presented in figure 2.3. The simulation model will be explained in more detail in section 3.
Fig. 2.3. Conceptual Model.
3. DESCRIPTION OF CONTROL STRATEGIES
3.1 Research Setting
This project is initiated by the UMCG Ambulancezorg. UMCG Ambulancezorg is responsible for the ambulance dispatches in Drenthe and a part of Friesland, the Netherlands. In table 3.1 an overview of the characteristics of UMCG Ambulancezorg is provided (region Drenthe). The goal is to reach the emergency location as quick as possible. Furthermore, it is important to reach this destination safely to protect its employees and other people on the road. Various regulations have been set in place to secure this safety. However, safety regulations might cause delay along the way. For instance, the regulation that ambulances must drive across an intersection at 20 km/h (5.5 m/s) if their traffic light is on red (Kobes, Ros, Groenewegen-ter Morsche, 2017). This
Accident Characteristics Conventional traffic characteristics iVRI characteristcs Input Simulation model Output Simulation Emergency Vehicle Characteristics - EV detection distance - Notification time period
Queue length at the moment the EV is detected
Time required to switch to green light
Time required to discharge the
queue
Safety Time Interval
Green time required for the approaching emergency vehicle
19
regulation is set in place decreases the risk of accidents as other vehicles might not have noticed the approaching ambulance.
The above section implies that safety regulations increase the response time of EMS. Recent innovations in traffic lights could enable a safer crossing at the intersection. Ultimately, the ambulance could cross the intersection at a higher speed and therefore reach the destination faster. This paper is dedicated to finding situations where the ambulance should be able to cross the intersection without decreasing its speed. Second, using this real-time strategy this study evaluates in what scenarios other approaches can have an extended green time to optimize traffic flow.
UMCG Ambulancezorg (Drenthe)
Number of ambulances 23
Number of locations 16
Number of dispatches per year 42,132
Number of A1 dispatches per year 20,827
Average Response Time 9:34 minutes
Average Time driving to location 6:55 minutes
Table 3.1. UMCG Ambulancezorg characteristics (2016).
3.2 Simulation Model
This paper adapts the model presented in Qin & Khan (2012). They introduce a real-time strategy that allows emergency vehicles to cross an intersection while minimizing traffic flow deterioration. This model is used to compare fixed time strategy to a real-time control strategy. Qin & Khan propose a model existing of two transitions, transition one is focused on letting the EV pass the intersection in a safe and quick manner, while transition two is focused on returning the operation of traffic light back to its normal operation. Due to time limitation, this paper will entirely focus on transition one as this is the priority of emergency vehicles.
3.2.1 Generation of vehicles
20
is assumed that vehicles follow a discrete random distribution. Using historical data, the number of vehicles arriving from each approach per hour the arrival distribution can be simulated. For each approach, the arrival of vehicles is generated. Further, the percentages of left and right turn and the operating speed of vehicles need to be calculated for the simulation. In the simulation of vehicle arrival, a random number between 0 and 1 will be generated. A number generated that is higher than 0.5 indicates a vehicle arrival for the specified approach.
3.2.2 Emergency vehicle distance
Subsequently to the generation of vehicle arrivals, queue lengths under normal operation can be calculated. This average queue length plus two times the standard deviation will be used to determine the emergency vehicle distance. Two times a standard deviation is added to ensure the emergency vehicle distance is sufficient in various situations. The model calculates a distance from the intersection where a detector should be placed to ensure a safe passage for the ambulance. As soon as an EV passes the detector, the traffic lights will be switched to green for the EV approach (Qin & Khan, 2012). Considering connected driving technologies, a fixed detection point is not expected to be placed near an intersection. Though, it is assumed that iVRI technology maintains real-time data, provided by GPS signals, considering speed and distance of the approaching EV. Additionally, a GPS signal might provide accurate data about the distance and speed of the approaching ambulance. It is presumed the traffic light control system will be able to determine the time necessary to switch green light to the approach with the approaching emergency vehicle. Therefore, the emergency vehicle distance will be used to calculate the time required to switch traffic lights to green before the EV arrives at the intersection. Also, this period should be sufficient to discharge any queue in front of the emergency vehicle. Besides, a safety time interval (STI) is added to prevent a collision. The STI is set to two seconds. The green time can be extended when the calculated period is not sufficient. Next to average queue length using historical data, also more extreme queue lengths will be considered under highly saturated traffic conditions.
3.2.3 Control strategies
21
historical data average signal plans are calculated and used for the fixed time control (FTC). For the simplicity of the calculation, the intersection analysed is simplified into a two-phase signal system. After the emergency vehicle distance is generated and the period required to provide a green light for the EV, a fixed time control will be simulated as the base case. The FTC operates as follows, when an EV is detected, the traffic light will allocate a fixed number of seconds of green light to the approach of the ambulance. As stated by IJedema (2018), the average fixed time provided for ambulance in the city of Groningen is 18 seconds. However, if this green time is not sufficient to clear the queue ahead of the EV, the emergency vehicle must decelerate or stop.
22 3.2.4 Intersection layout
For the simulation, an intersection is chosen in the region near the UMCG since ambulances of UMCG Ambulancezorg often cross this intersection. The intersection is depicted in figure 3.1. To simplify the calculation, merely conventional vehicles and emergency vehicles will be considered. Bicycles, buses and pedestrians are excluded from the simulation due to time limitation. Therefore, the intersection can be considered as a simple four-leg intersection. When referring to the intersection the following approaches are considered. Approach 1 corresponds to the Europaweg towards the UMCG. Approach 2 is the Griffeweg towards the Sontweg. Approach 3 corresponds with the Europaweg towards the Europaplein and Approach 4 resembles the Sontweg towards the Griffeweg.
Fig. 3.1. Intersection Groningen.
3.3 Data Collection
23
These historical data correspond to intersection introduced in the previous section. The data collected is from the date 20 March 2018. A limitation of this study is the amount of data as average volumes of vehicle arrivals for various days during a week or month can differ. Therefore, circumstances and their influence on the FTC and RTC strategies cannot be compared. Simulations will be executed using the collected data. However, this will be compared to estimated data to simulate more dense traffic situations. Data provided by Gemeente Groningen is analysed before using it for the simulation. First, the fixed timing plan according to Webster’s method as described in (Qin & Khan, 2012) was calculated. Second, the fixed timing plan using the historical data provided by Gemeente Groningen was calculated. Even though results are similar, the historical data is used as this applies to the specific intersection analysed. The results of the fixed signal timing plan are shown in table 3.2. Where the red time period consists of one second. This red time indicates the period where all traffic lights are on red light to clear the intersection.
Green phase 1 Green phase 2 Yellow Red Cycle time
34 seconds 26 seconds 3 seconds 1 second 68 seconds
Table 3.2. Signal timing plan fixed control.
Supplementary, average arrival of vehicles per hour and their speed is calculated. Last, from the data provided by Gemeente Groningen, average right and average left turn are determined. Given the exact volumes on each lane, the average right and left turn can be calculated as each lane indicates one specific direction. An overview can be found in Appendix A. An overview of conventional vehicle characteristics is provided in table 3.3. It can be noted that the average vehicle arrival at approach 1 and 3 is considerably higher than approach 2 and 4. This can be explained by the fact that these approaches come and go towards a high road.
Approach Average amount of vehicles per approach (veh/h)
Average Speed (m/s) Average turn right Average turn left 1 914.50 14.80 18% 18% 2 431.92 10.80 35% 29% 3 896.08 9.72 21% 11% 4 418.17 11.47 23% 34%
24
As presented in figure 2.3, the following characteristics will be used as input elements for the simulation model. According to the IFV accident report (2017) in 79% of the accidents with priority vehicles at an intersection, the priority vehicle crossed the speed limit of 20 km/h (e.g. 5.50 m/s). On average the registered speed of the priority vehicle was 46 km/h (e.g. 12.78). Therefore, in the analysis, it will be tested if an emergency vehicle can cross the intersection at an operating speed of 15 m/s. Besides, as indicated by UMCG Ambulancezorg (2018) to reach an accident location or hospital as soon as possible, the ultimate goals would be to cross an intersection at operating speed. Therefore, it will be analysed if an EV can cross the intersection at the operating speed of 20 m/s. Last, as indicated by the developers of iVRI software, in the near future the software of an iVRI will be able to look and plan approximately 300 metres ahead. In the simulation, the implications of this distance will be evaluated.
4. SIMULATION
The above-presented model will be simulated using software MATLAB (The MathWorks Inc., 2003). For all simulations, 20 cycles are simulated. The simulations described below use a set of predetermined parameters as described in Appendix A. First, a simulation is executed with the EV arriving from approach 1. Subsequently, a simulation is carried out using a high density of vehicle arrivals. Last, a simulation is performed with the EV arriving from approach 2.
25
end of the transition. Last, under RTC strategy green time can be allocated to the side road. The longer the green light allocation, the less the EV interrupts the traffic flow.
4.1 Analysis using Historical Data
4.1.1 Queue length under normal operation
First, a normal operation is simulated to generate queue lengths at each approach. The average queue length plus two times the standard deviation is used to ensure that under most traffic situations the emergency vehicle distance will suffice. The average queue length under normal operation with two times the standard deviation on approach 1, 2, 3 and 4 are respectively 8.39, 3.20, 10.07 and 3.07 vehicles. The overall results are provided in Appendix B. Using the above found queue lengths on all approaches the following emergency vehicle detection (EVD) distances are simulated. An overview of the results is shown in table 4.1. These distances should be sufficient for an emergency vehicle to pass the intersection without delay. In the subsequent simulations, tests are performed if this applies in several situations. The EVD distance of 300 metres is used as iVRI software is likely to adjust the traffic light based on information received 300 metres in advance, as described in section 2.2.2. Under scenarios with an EVD distance larger than 300 metres, software should be expanded. For the scenario of 300 metres it will be analysed at what operating speed the emergency vehicle will be able to cross the intersection without decreasing its speed. However, as indicated in table 4.1, at an operating speed of 15 m/s, the EVD distance should already be larger than 300 metres. Further, a simulation will be performed for an EVD distance of 500 metres, as this distance should suffice in most situations as observed in table 4.1. Next, the results under EVD distance of more extreme cases as 100 and 900 metres will be simulated to gain insight into these more exceptional scenarios.
EV arriving from
Approach 1 Approach 2 Approach 3 Approach 4 EVD Distance (metre)
EV speed = 15 m/s
362.00 221.47 407.16 217.97
EVD Distance (metre) EV speed = 20 m/s
482.00 295.32 542.88 290.63
EVD period (seconds) 24.00 14.77 27.14 14.53
26 4.1.2 Simulation of fixed time control
After determining the minimum EV detection distance, the base scenario of fixed time control can be simulated. Using the predetermined signal timing plan describe in section 3.3, the following scenarios of an EV entering the control system were simulated as indicated by Qin & Khan (2012): EV enters at the beginning of green time, at the middle of green time, at the beginning of yellow light, at the beginning of red light of the main road. Moreover, the scenarios of the emergency vehicle entering the system at the beginning and middle of green light for the side road are considered. The scenarios are summarised in table 4.2.
Scenario Entry moment Time
1 At the beginning of green time (main road) t = 0 s
2 Middle of green light (main road) t = 17 s
3 The beginning of yellow light (main road) t = 34 s 4 The beginning or red light (main road) t = 37 s
5 Beginning of green light (side road) t = 38 s
6 Middle of green light (side road) t = 55 s
Table 4.2. Emergency vehicle arrival scenarios.
The FTC strategy is simulated using a fixed period of 10 seconds. Each scenario is simulated 6 cycles. An overview of each scenario is provided in table 4.3. The elaborate results can be found in Appendix B. It can be noted that the average spare time is considerably high as the fixed period is 10 seconds. Solely under scenario 1, the strategy failed to let the EV pass at operating speed. Looking at table 4.3, it can be observed that the queue length at approach 3 and 4 are quite low, while the average queue length at approach 1 and 2 are somewhat higher. However, no extreme queue lengths are found under the fixed time control strategy with a fixed time set to 10 seconds.
Scenario More time needed (s) Spare time (s) Queue on approach
1 2 3 4 1 2.00 3.78 1.17 0.67 2.50 0.33 2 0.00 10.00 1.00 1.33 0.00 0.50 3 0.00 9.70 1.67 2.00 0.00 0.83 4 0.00 8.98 1.67 2.00 0.00 0.83 5 0.00 8.26 1.17 2.17 0.00 1.00 6 0.00 5.50 1.00 0.33 1.17 0.33
27
The same simulation is also executed using a fixed time of t = 20 seconds, t = 40 seconds and t = 50 seconds. A summary of these results is shown in table 4.4. Note, the increase in spare time as the fixed time is set to a higher value. However, the number of failures to provide green light decreases. The queue lengths at different approaches suggest the same trend, where the queue lengths at approach 1 and 2 are higher than at approach 3 and 4.
Amount of failures Spare time Queue on approach
1 2 3 4
t = 10 2 7.94 1.23 1.42 0.61 0.64
t = 20 0 17.38 2.31 1.72 0.44 0.92
t = 40 0 37.38 4.61 2.53 0.17 1.31
t = 50 0 47.38 5.64 2.78 0.28 1.56
Table 4.4. Average queue length under fixed time control.
4.1.3 Simulation of real-time control
For the simulation of real-time traffic control, the six scenarios are the same as described for the fixed time control strategy. In the table below the results of real-time traffic control are shown. This simulation is executed with the traffic volume based on historical data and a detection distance of 300 metres. As shown in table 4.5, the average green time allocated to the side road is quite low (e.g. less than 10 seconds), while the average spare time is 44% of the time. As 300 metres is comparable to the 20 seconds FTC strategy, it appears the average spare time is less than under the FTC strategy. This could indicate a more efficient operation.
Scenario Green time for the side road
More time needed Spare time Queue on approach 1 2 3 4 1 0.00 0.00 6.85 1.50 0.83 0.33 0.66 2 0.00 0.00 11.00 1.83 1.50 0.00 1.17 3 0.00 0.00 10.70 2.00 2.00 0.00 1.33 4 5.00 0.00 5.98 1.83 0.66 1.33 0.83 5 3.67 0.00 8.27 1.50 1.50 1.00 0.66 6 0.00 0.00 10.52 1.83 0.66 0.00 0.17
Table 4.5. Average queue length under real-time control (speed EV = 15 m/s, L = 300 m).
28
seconds. A summary of the results is shown in table 4.6. For the simulation results of EV = 20 m/s, please refer to Appendix B. As can be observed, when the EVD distance exceeds 300 metres, the green time allocated to the side road increases significantly. Whereas under EVD distances of 100 and 300 metres the amount of failures is 20 and 2 respectively, the number of failures of the real-time system diminishes to zero with an EVD distance longer than 300 metres. Next, the spare time decreases significantly when the EVD distance is longer than 300 metres. Though, the average queue lengths on approach 1 and 3 start to increase with the EVD distance set to 500 and 900 metres.
Distance Failures Green Time side road More time needed Spare time Queue on approach 1 2 3 4 L = 100 20 0.00 3.32 0.00 1.42 1.14 1.92 0.69 L = 300 2 1.44 0.00 8.89 1.75 1.19 0.44 0.81 L = 500 0 18.60 0.00 1.62 3.33 0.36 3.69 0.19 L = 900 0 45.27 0.00 1.62 6.22 0.39 6.92 0.19
Table 4.6. Average queue length under Real-Time Control (Speed EV = 15 m/s).
4.2 Analysis with Highly Dense Traffic
29
Fig. 4.1. Average queue length normal operation under different saturation rates.
Under highly dense traffic situations, the fixed time traffic control provides comparable results to the simulations with historical data. It can be noted that under highly dense traffic situations the average queue length at each approach increases significantly which indicates a less efficient operation. After performing the simulation under fixed time control, also the RTC strategy is simulated for the highly dense traffic situation. Under real-time control, the results show the same trend as presented in section 4.1. The emergency vehicle can cross the intersection with an EVD distance larger than 300 metres. Furthermore, considering this distance, the RTC system can allocate considerable green time (e.g. more than 10 seconds) to other approaches. Under the EVD of 500 and 900 metres respectively 45 and 80% of the real-time strategy is allocated as a green light to the side road. Where this period could be considered as a waste of time under the FTC strategy, applying the RTC strategy, the system becomes more efficient. Please refer to Appendix C for an overview of the results.
4.3 Emergency Vehicle arriving from Side Road
Last, to understand the benefits of the real-time control strategy, this section provides the results of instances where the EV arrives at the intersection from approach two. In table 4.7 an overview of the results under FTC is provided. As can be seen, the spare time is considerably high. The queues formed at approach 1 and 3 are markedly larger than the queues at approach 2 and 4.
0 20 40 60 80 100 120 140 160 180 V = 500 V = 1000 V = 1500 V = 2000 V= 2500
Average Queue Length
30
Spare Time Approach 1 Approach 2 Approach 3 Approach 4
t = 10 8.63 s 3.64 0.36 4.64 0.03
t = 20 18.63 s 4.67 0.67 5.86 0.06
t = 40 38.63 s 7.08 1.53 8.28 0.00
t = 50 48.63 s 8.25 1.75 9.94 0.06
Table 4.7. Simulation results of fixed time control (EV from side road).
When comparing simulating the real-time control strategy, the outcomes confirm those found in section 4.1. The green time for the approaches apart from approach 2 increase significantly as the EVD distances become larger than 300 metres. Besides, the average spare time decreases noticeably with an EV detection distance larger than 300 metres. In fact, the average queue lengths at all approaches are at first sight, lower than under the fixed time control strategy. For the results of the real-time control strategy with an EV approaching from the side road (approach 2), please refer to Appendix E. Like the results of section 4.1, the green time for other approaches increases significantly with an EVD distance larger than 300 metres. Besides, the average spare time decreases noticeably with an EV detection distance larger than 300 metres. In fact, the average queue lengths at all approaches are, at first sight, lower than under the fixed time control. Please refer to Appendix E for an overview of the results.
5. DISCUSSION
5.1 Comparison Control Strategies EV from Main Road
31
Fig. 5.1. Average spare time.
To analyse this, the average green time under different emergency vehicle detection distances is evaluated. It is confirmed that the longer the EVD distance, the more time is allocated to the green side. Typically, with an EVD distance larger than 300 metres, almost all spare time under FTC is allocated as green light under the RTC. This indicates that the inefficiency under the fixed time control is allocated as green light for other approaches under the real-time control strategy. Thus, the efficiency increases under the real-time control strategy. Please refer to Appendix E for an indication of this allocation. Finally, the average queue length under both strategies is compared. As can be noticed in figure 5.2, the average queue length on approach 1 and 3 increases significantly under the RTC. This could indicate a less efficient operation on the intersection as the queue length is one of the performance indicators of the intersections efficiency.
Fig. 5.2. Average queue length fixed time compared to real-time.
However, when comparing the overall average queue lengths, the overall queue length indeed increases under real-time control strategy of t = 10 seconds and t = 50 seconds (e.g. L = 100 m and L = 900 m). The overall average queue length decreases under an EV
0 10 20 30 40 50 t = 10 s t = 20 s t = 40 s t = 50 s
Average Spare Time
Fixed Real 0 2 4 6 8 10 12 14
Approach 1 Approach 2 Approach 3 Approach 4
Average Queue Length
32
detection distance of 300 and 500 metres (e.g. t = 20 seconds and t = 40 seconds). In fact, the overall queue length under EVD distance of 300 and 500 metres decrease respectively 28% and 14%. With the decrease in overall queue length in combination with time allocated to other approaches and less spare time, it could be stated that the RTC system ensures a smoother and therefore more efficient operation on the intersection. This statement holds especially for an emergency vehicle detection distance between 300 and 500 metres. Also, when the EVD distance is larger than 300 metres, the real-time control strategy has a 100% success rate, so the RTC strategy can increase the safety of ambulances at intersection.
5.2 Comparison Control Strategies EV from Side Road
When comparing the simulation results of the fixed and real-time control strategies for the EV arriving from approach 2, comparable results as described above can be found. Under the RTC strategy, the average spare time is considerably lower than under the fixed time control strategy. This indicates a more efficient operation at the intersection.
Fig. 5.3. Average spare time with EV from approach 2.
Also, average queue lengths under both strategies are compared. As depicted in figure 5.4 the overall average queue lengths under the real-time control are lower. Merely at approach 4, the average queue length increases under the RTC strategy. This indicates more delay at approach 4 under the real-time strategy than under the fixed control strategy. Although this is the case, when comparing the overall queue lengths under both strategies, it is observed that the overall average queue length at the intersection decreases significantly. Respectively under t = 10, t = 20, t =4 and t = 50, the overall average queue length decreases with 102,60%, 191,37%, 360,61% and 234,88%. This shows that overall the real-time strategy is significantly more efficient. While ensuring a safe and quick passage of the intersection for the EV, the traffic flow disruption is minimized.
0 20 40 60
t = 10 s t = 20 s t = 40 s t = 50 s
Average Spare Time
33
Fig. 5.4. Average queue length with EV arriving from approach 2.
5.3 Value of Real-Time Control Strategies
It can be argued that a real-time control strategy provides promising results for safe passing for ambulances at intersections. Typically, the amount of failures decreases to zero when the EVD distance is larger than 300 meters, so the ambulance is certain to receive green light under this strategy. This means the EV can cross the intersection at operating speed. This holds for the EV driving at 15 m/s and 20 m/s. Thus, under higher emergency situations (category A1 dispatches), the ambulance can be assured of a safe crossing when the traffic light starts to adjust the lights before the EV is 300 metres away from the intersection. When considering an estimated time of arrival instead of an emergency vehicle detection distance, the time period should be set within in a range between 20 and 35 seconds to assure a safe passage for the ambulance.
On the other hand, the amount of failures under fixed time control is smaller than under the RTC strategy if predefined fixed times are set to 10 and 20 seconds respectively. As stated in section 3.2.3, the current control strategy on average has a fixed time period of 18 seconds. Although under the current fixed time control strategy the EV is often able to pass the intersection without delay, the smooth traffic flow is highly disturbed by the fixed time control as can be seen by the high values of the spare time and the generated queue lengths. Therefore, if the emergency vehicle wants to be assured of a safe passage without disturbing the traffic flow, a real-time control strategy can be applied. Further, current regulations withhold an emergency vehicle to cross the intersection at operating speed. If the real-time control strategy is proven to be fully reliable, this regulation could in the future be removed and allow an EV to cross the intersection at operating speed.
0 2 4 6 8 10 12
Approach 1 Approach 2 Approach 3 Approach 4
Average Queue Length
34
This simulation is performed with a minimum spare time of 7 seconds before the green light will be allocated to other approaches. This is set as a human driver can be expected to be able to react to changes in traffic light within 5 seconds. However, as described by Guler et al. (2014), when more autonomous vehicles are driving in the city of Groningen, this minimum green time could eventually be decreased. This because autonomous vehicles are able to react faster than human drivers. Therefore, the real-time control strategy could optimise the efficiency of an intersection when the penetration rate of autonomous driving increases.
35
6. CONCLUSIONS
This paper addresses the applicability of a real-time control strategy for the city of Groningen. The goal is, while adapting a real-time control strategy, to gain insight into the safety and smoothness of Emergency Medical Services crossing an intersection. It is studied if and under what circumstances an emergency vehicle can cross the intersection at operating speed. The second goal is to evaluate if the emergency vehicle can cross the intersection while minimising the interruption of the conventional traffic flow. This effect is evaluated by comparing the real-time control strategy to a fixed time control strategy. The developed simulation addresses various operating speed of ambulances, emergency vehicle detection distances and conventional traffic densities. The simulations of the real-time control strategies show that real-real-time control strategy ensures a safe passage for an ambulance by providing a green light. This strategy is optimal with distances between 300 and 500 metres. When the EVD distance is set between these parameters, spare time can be optimally allocated to other approaches to decrease the overall delay at these approaches. Therefore, it can be concluded that the real-time strategy can enhance the safety of an emergency vehicle while minimizing the total delay at the intersection.
The results of this work provide promising initial insight into increasing the safety of emergency vehicles in the city of Groningen. However, the initial software to be implemented in traffic lights in the city of Groningen are constrained to 300 metres. Therefore, the software requires an extension before the evaluated real-time strategy can be optimised in the city. While the real-time strategy allows often a safe passage for ambulances with a detection distance smaller than 300 metres, this would decrease the smooth traffic flow at the intersection.
The applicability of the strategy is limited to other cities. To provide a grounded conclusion about the applicability to other cities and intersections, the simulation should be repeated with the corresponding data of that city and intersection. However, it can be noted, in cities with similar traffic situation, the results will most likely be similar. Only under highly dense traffic situations (e.g. more than 2000 vehicles per hour) this strategy is questionable and requires more research. This might be the case for cities considerably larger than Groningen.
36
time of an ambulance can be decreased, as it can cross the intersection at operating speed. Less deceleration and accelerations can enhance the comfort of the patient. Last, a real-time strategy can minimize the total delay at the intersection while adhering to priority requests.
Theoretical implication of this research is the applicability of a real-time control strategy in the Netherlands. Whereas the strategy previously has been tested with estimated data corresponding to Canada, this paper has addressed the applicability of the real-time strategy to a city in the Netherlands, given historical data of the city of Groningen. It adds to the existing field of research by providing insight into the enhancement of safety of ambulances. This is next to the abundant strategies to decrease the response times of emergency vehicles. Besides insight into the safety of emergency vehicles, it provides initial insight into a real-time traffic control system that can increase safety while adhering priority to emergency vehicles.
For management in the field, this research provides initial insight into the parameters under which a real-time control strategy could result in an optimal operation at the intersection while adhering green light to an approaching emergency vehicle. This optimal situation could be reached when an EV detection distance is set between 300 and 500 metres. With these distances, the average delay can be minimized while assuring a safe passage for the ambulance. Though the soon to be implemented traffic lights contain a software that is initially not able to adjust its lights based on information further than 300 metres ahead of the intersection. This indicates that software might need further improvements before the priority vehicle can cross an intersection at its operating speed while minimizing traffic flow deterioration.
Limitations of this study are the amount of vehicle data collected as a more extensive dataset could provide parameters that ensure a safe passage for an ambulance under various traffic situations. Additionally, as transport modes as busses, bicycles, pedestrians and other priority vehicles are excluded from this study, the provided results do not match the reality.
37 REFERENCES
Aboueljinane, L., Sahin, E., Jemai, Z. 2013. A review on simulation models applied to emergency medical service operations. Computers & Industrial Engineering, 66: 734-750.
Aboueljinane, L., Sahin, E., Jemai, Z., Marty, J. 2014. A simulation study to improve the performance of an emergency medical service: Application to the French Val-de-Marne department. Simulation Modelling Practice and Theory, 47: 46-59.
Ambulancezorg Nederland. 16 November 2017. Tabellenboek 2016, Zwolle.
Ambulancezorg Nederland. 2013. Uniform Begrippenkader Ambulancezorg, Zwolle.
Aringhieri, R., Carello, G., Morale, D. 2013. Supporting decision making to improve the performance of an Italian emergency medical service. Annals of Operations Research, 236(1): 131-148.
Aringhieri, R. Bruni, M.E., Khodaparasti, S., van Essen, J.T. 2017. Emergency medical service and beyond: Addressing new challenges through a wide literature review. Computers and Operations Research, 78: 349-368.
Bos, N., Krol, M., Veenvliet, C., Plass, A.M. 2015. Ambulance care in Europe: Organization and practices of ambulance services in 14 European countries. Utrecht: Nivel.
Burke, C.S., Salas, E., Kincaid, J.P. 2001. Emergency vehicles that become accidents statistics: Understanding and limiting accidents involving emergency vehicles. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 45(4): 508-512.
Chiu, P., Lin, C., Wu, C., Fang, P., Lu, C., Hsu, H., Chi, C. 2018. Ambulance traffic accidents in Taiwan, Journal of the Formosan Medical Association, 1-9.
Donabedian, A. 1980. Definition of Quality and Approaches to its Assessment: explorations in quality assessment and monitoring. Health Administration Press.
38
Fassbender, K., Balucani, C., Walter, S., Levine, S.R., Haass, A., Grotta, J. 2015. Streamlining of prehospital stroke management: the golden hour, Lancet Neurol, 12: 585-596.
Feng, Y., Head, K.L., Khoshmagham, S., Zamanipour, M. 2015. A real-time signal control in a connected vehicle environment, Transportation Research Part C, 55: 460-473.
Goldberg, J., Dietrich, R., Chen, J.M., Mitwas, M., Valenzuela, T., Chriss, E. 1990. A simulation model for evaluating a set of emergency vehicle base locations: development, validation and usage. Socio-Economic Planning Sciences, 24(2): 125-141.
Guler, S.I., Menendez, M., Meier, L. 2014. Using connected vehicle technology to improve the efficiency of intersections, Transportation Research Part C, 46: 121-131.
He, Q., Head, K.L. Ding, J. 2012. PAMSCOD: platoon-based arterial multi-modal signal control with online data. Transportation Research Part C, 20(1): 164-184.
He, Q., Head, K.L., Ding, J. 2014. Multi-modal traffic signal control with priority, signal actuation and coordination, Transportation Research Part C, 46: 1-17.
Heida, J., Hoendervanger, J., Veer, L. 2015. Ordening ambulancezorg. Den Haag: Strategies in Regulated Markets.
IJedema, W. 2018. Interview with Urban Development specialist. Gemeente Groningen. Meeting at 5 April 2018.
Jin, J., Ma, X., Kosonen, I. 2017. An intelligent control system for traffic lights with simulation-based evaluation, Control engineering Practice, 58: 24-33.
Kale, S.B., Dhok, G.P. 2013. Design of ambulance and traffic control. International Journal of Innovative Technology and Exploring Engineering, 2(5): 211-214.
Kobes, M., Ros, A., Groenewegen-ter Morsche, K. 2017. Ongevallenstatistiek voorrangsvoertuigen 2014-2015. Instituut Fysieke Veiligheid.
Krishna, A., Hartha, B.A., Nair, V.S. 2017. Dynamic traffic light system for unhindered passing of high priority vehicles. Global Humanitarian Technology Conference.
39
Linganagouda, R., Raju, P., Patil, A. 2016. Automatic intelligent traffic control system. International Journal of Advanced Research in Electrical, Electronics and Instrumentation Engineering, 5(7): 5902-5906.
McCormack, R., Coates, G. 2015. A simulation model to enable the optimization of ambulance fleet allocation and base station location for increased patient survival. European Journal of Operations Research, 247(1): 294-209.
McLay, L.A., Mayorga, M.E. 2010. Evaluating emergency medical service performance measures. Health Care Management Science, 13(2): 124-136.
Meulenaere, I., Kessels, B. 2018. Meeting with solution architect of Ericsson. Ericsson. Meeting at 19 February 2018.
Missikpode, C., Peek-Asa, C., Young, T., Hamman, C. 2018. Does crash risk increase when emergency vehicles are driving with lights and sirens? Accident Analysis and Prevention, 113: 257-252.
Nellore, K., Hancke, G.P. 2016. Traffic management for emergency vehicle priority based on visual sensing. Sensors: Switzerland, 16(11): 1-22.
Peleg, K., 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.
Pinto, L.R., Silva, P.M.S., Young, T.P. 2015. A generic method to develop simulation models for ambulance systems. Simulation Modelling Practice and Theory, 51(2): 170-183.
Pons, P.T., Haukoos, J.S., Bludworth, W., Cribley, T., Pons, K.A., Markovchick, V.J. 2005. Paramedic response time: does it affect patient survival? Academy Emergency Medicine, 17(9): 594-600.
Process Ambulance; Ambulancezorg Nederland; www.ambulancezorg.nl ; Retrieved at 21/02/2018.
Qin, X. 2005. Real-time control and optimization of traffic signal timing transition for emergency vehicle pre-emption. Unpublished doctoral dissertation, Carleton University, Ottawa, Ontario.
40
Reuter-Oppermann, M., van den Berg, P.L., Vile, J.L 2017. Logistics for emergency medical service systems. Health Systems.
Sanddal, T.L., Sanddal, N.D., Ward, N., Stanley, L. 2010. Ambulance crash characteristics in the US defined by the popular press: a retrospective analysis. Emergency Medicine International.
Slattery, D.E., Silver, A. 2009. The hazards of providing care in emergency vehicles: an opportunity for reform. Prehospital Emergency Care, 13(3): 388-397.
Stein, C., Wallis, L., Adetunji, O. 2015. The effect of emergency medical services vehicle location and response strategy on response times. South African Journal of Industrial Engineering, 26(2): 26-40.
Intelligente Verkeersregelsinstallaties; Talking Traffic; Platform Beter Benutten;
https://www.beterbenutten.nl ; retrieved at 25/03/18.
Yang, K., Guler, S.I., Menendez, M. 2016. Isolated intersection control for various levels of vehicle technology: conventional, connected and automated vehicles. Transportation Research Part C, 72: 109-129.
41
APPENDIX A INPUT SIMULATION
Table A.1. Average Time Traffic Light
Approach Average time Green (seconds) Average time Yellow (seconds) Average time Red (seconds) 1 19.20 03.00 60.80 2 13.20 03.00 69.50 3 15.60 03.00 56.90 4 12.80 02.90 67.00
Table A.2. Parameters for Simulation
Parameter Value
Saturation 2000 vehicles/hour
Lost time per phase 3 seconds
All red time 2 seconds
Lost time per cycle 8 seconds
42
APPENDIX B RESULTS EMERGENCY VEHICLE FROM MAIN ROAD
Table B.1 Average queue length under normal operation.
Cycle Approach 1 Approach 2 Approach 3 Approach 4
43
Table B.2. Results of Fixed Time control strategy (speed EV = 15 m/s, L = 300 m).
Scenario Cycle More time needed (s) Spare time (s) Queue on approach
44 6 5 3.88 1 0 1 1 6 4.60 0 0 2 0 7 7.48 1 1 1 1 8 4.24 0 1 0 0 9 6.40 2 0 1 0 10 6.40 2 0 2 0 Average 2.00 7.94 1.23 1.42 0.61 0.65
Table B.3. Results of Real-Time control strategy (speed EV = 15 m/s, L = 300 m).
Scenario Cycle Green time for side road
45 8 10 0 3 0 2 0 9 10 0 2 0 2 0 10 0 12.2 1 1 0 2 5 5 0 12.5 2 3 0 0 6 0 11.4 1 1 0 2 7 0 12.5 0 2 0 1 8 11 0 3 0 3 0 9 11 0 2 0 3 0 10 0 13.2 1 3 0 1 6 5 0 8.9 0 0 0 0 6 0 9.6 2 0 0 0 7 0 12.5 0 1 0 0 8 0 9,3 3 1 0 1 9 0 11.4 4 1 0 0 10 0 11.4 2 1 0 0 Average 1.44 0 8.89 1.7 5 1.1 9 0.4 4 0.8 1
Table B.4. Average spare time and queue length under EV = 20m/s.
46
APPENDIX C RESULTS HIGHLY DENSE TRAFFIC
Table C.1. Average queue length under normal operation.
Approach 1 Approach 2 Approach 3 Approach 4
V = 500 4.00 3.00 5.20 3.30 V = 1000 8.10 5.40 11.60 5.40 V = 1500 11.90 6.80 14.80 8.00 V = 2000 59.40 10.70 101.10 10.70 V= 2500 129.70 13.20 153.70 18.10
Fig. C.1. Time Needed under fixed time control.
0 20 40 60 80 100 V = 500 V = 1000 V = 1500 V = 2000 V= 2500
More Time Needed
47
Fig. C.2. Average queue length under fixed time control.
Table C.2. Results under fixed time control.
More time needed (s) Spare time (s) Queue on approach
1 2 3 4
t = 10 3.21 7.38 2.00 2.75 2.92 4.11
t = 20 0.00 14.44 4.33 3.58 1.69 5.17
t = 40 0.00 34.44 8.78 5.00 0.56 7.31
t = 50 0.00 44.44 11.31 5.83 0.36 8.42
Table C.3. Results under real-time control.
Green Time side road (s) More time needed (s) Spare time (s) Approach 1 2 3 4 L = 100 0.00 6.60 1.10 2.78 2.14 4.58 2.83 L =300 1.58 2.64 6.73 3.00 2.75 1.31 3.47 L = 500 14.73 0.00 3.11 6.19 1.03 5.78 1.50 L = 900 42.49 0.00 1.46 12.50 1.08 12.33 1.31 0 10 20 30 40 v= 500 V = 1000 V = 1500 V = 2000 V= 2500
Overall Average Queue Length
48
APPENDIX D RESULTS WITH EMERGENCY VEHICLE FROM SIDE ROAD
Table D.1. Simulation results of real-time control (EV from side road).
Green time for side road Spare time Queue on approach
1 2 3 4
L = 100 0.00 1.30 1.39 0.64 1.31 0.94
L = 300 2.94 8.08 1.42 0.53 1.42 0.50
L = 500 19.90 1.57 0.42 1.56 0.50 1.69
49
APPENDIX E COMPARISON CONTROL STRATEGIES
Fig. E.1. Average Green time Allocated with EV from approach 1.
Fig. E.2. Spare Time and Green Time Allocation with EV from approach 1.
Fig. E.3. Overall Average Queue length with EV from approach 1. 0 10 20 30 40 50 t = 10 s t = 20 s t = 40 s t = 50 s
Average Green Time for Side Road
0 10 20 30 40 50 t = 10 s t = 20 s t = 40 s t = 50 s
Spare Time and Green Time Allocation
Spare time (Fixed) Green Time (Real)
0 1 2 3 4 t = 10 t = 20 t = 40 t = 50
Overall Average Queue Length
50
Fig. E.4. Average Green time allocated with EV from approach 2.
Fig. E.5. Spare Time and Green Time allocation with EV from approach 2.
Fig. E.6. Overall Average Queue Length with EV approaching at approach 2. 0 10 20 30 40 50 t = 10 s t = 20 s t = 40 s t = 50 s
Average Green time for side road
0 10 20 30 40 50 60 t = 10 s t = 20 s t = 40 s t = 50 s
Spare Time and Green Time Allocation
Spare time (Fixed) Green Time (Real)
0 1 2 3 4 5 6 t = 10 t = 20 t = 40 t = 50
