1. Introduction
1.1. Background
According to the International Panel on Climate Change (IPCC) 2007 report [1], climate change has been happening for many decades. One of the most significant discussions throughout the world is the consequences of climate change because it is harmful to human beings, for example, global warming, sea levels rising, acid rain, flooding, extreme precipitation and drought, etc. Greenhouse gases (GHG) have widely been recognised as an important component contributing to climate change; the increasing concentration of gases in the atmosphere has gradually created the greenhouse effect. In fact, according to the World Resources Institute, 81% of GHG are from Carbon dioxide (CO2) in developed countries and 41% are from developing countries. Therefore, carbon emissions have been thought to be key factors contributing to GHG. In the UK, according to the Department of Energy and Climate Change, 26% of CO2 emissions were from the transport sector in 2011, which is around 119 million tons of CO2. So, it is becoming increasingly difficult to ignore carbon emissions from the transport sector.
According to the Department of Transport, the UK has been encouraging the public to cycle as one of its sustainable travel policies to reduce CO2 emissions, as cycling is considered to be pollution free and is a good alternative to driving a car to commute to work. However, a previous study [2] reported that people will only have the incentive to cycle when they find cycling to be safe and convenient. Countries such as Denmark, Germany and the Netherlands provide a comprehensive integration for cyclists to travel around cities. Bicycle facilities such as separate cycling paths alongside traffic roads, connections with public transport and cycling education are heavily utilised in these countries. Therefore, the general public have the incentive to cycle and effectively reduce carbon emissions. On the other hand, countries with less integrated cycling facilities such as lanes being shared between cars and bicycles, not only discourage the public from cycling, but could possibly increase CO2 emissions. The major problem of having shared road lanes is that cars have to reduce their speed to avoid hitting bicycles, or they have to increase their speed to overtake the bicycles. A recent study from Cappiello [3] suggests that acceleration from cars significantly increases carbon emissions. Cars are very often accelerating/decelerating when they share lanes with bicycles; consequently, this project seeks to explain the development of simulating the interaction between cars and bicycles, and then to see if bicycles have a positive or negative impact on the carbon emissions from the cars.
1.2. Aim and objective
The main objective of this project is to develop a simulation tool between bicycles and cars. Therefore, cars and bicycles should be running freely on the road. Moreover, while they are running on the road, some traffic behaviours should be applied to ensure the simulation is realistic and accurate. In order to see the impact of carbon emissions, this project also needs to develop a method in order to gain the data from the simulation tool and apply it to a traffic emissions model. Moreover, the ultimate goal of this project is to see the impact that cycling has on carbon emissions from cars; a few scenarios with different sizes of road network, different proportions of both vehicles and different traffic behaviours should be approached and tested too. The end result of this project is to find out if cycling really helps to reduce carbon emissions in different situations, and consequently to help the government to refine the cycling infrastructure of lanes shared between bicycles and cars.
To achieve the objective a few steps have to be taken. Firstly, to review the published mixed vehicle simulation models and it gain a thorough understanding of the interaction between bicycles and cars. This could help the project to observe the characteristics of bicycles and cars in order to build the simulation model. Although the project is interested in bicycles and cars, it is also necessary to observe the related traffic behaviours such as overtaking, traffic lights or junction exits in order to make it as realistic as possible. After defining the interaction between bicycles and cars, a clear idea of what the program should achieve will be seen, such as designing and identifying the UML classes for the program and also the work flow for the program. Secondly, by reviewing the published traffic emissions models this will help to understand what kinds of data this project would need in order to output the carbon emissions. Traffic emissions models are very complex; in fact, traffic emissions models are expected to involve many aspects, such as the speed of the car, the acceleration rate of the car, the size of the engine of the car, the fuel type of the car and the weight of the car, etc. Therefore, collecting the right and precise data will help the model to reduce the complexity while obtaining a fair result. Thirdly, after knowing how the program is approached and what kinds of data the project will need, the next step is to design the different kinds of scenarios in order to obtain effective results. Lastly, statistical approaches are taken in order to analyse and compare the results. The program is computed by Java.
1.3. Limitations of the study
So far there has been little discussion about the development of a mixed vehicle traffic emissions model. The world’s first bicycle-vehicle model was developed by Faghfri and Eghyhaziova in 1999 [4]. The paper provides a microscopic simulation model to study the interaction between the bicycles and cars. Although it gives good direction of what the program should be like, it only focuses on high levels of traffic behaviours such as lane-changing, the impact of pedestrians, the impact of bus stops and car parks, etc., and these behaviours are far too complex for this model. The model also does not indicate how overtaking behaviour is performed. However, it provides a car-following model to avoid cars crashing in the model. There are plenty of other published articles that have been developing simulation models of the interaction between bicycles and cars. However, most of them are interested in road accidents, safety issues or congestion. There are a few models researching traffic emissions models, but most of them only study vehicles without including bicycles. Therefore, there is no existing study to address the problem this project aims to uncover. Ultimately, the main purpose of this study is to develop a simulation model, from which bicycles and cars will react differently depending on the traffic conditions. While the cars are overtaking, all the data, such as the time, position or acceleration, are recorded. Later on, the data will be put into a traffic emissions model to produce the carbon emissions. Due to time constraints, this paper cannot provide a comprehensive study of traffic behaviours running on a real-sized city network. The reader should bear in mind that this paper only provides a small sample of bicycles and cars running on roads with limited traffic behaviours.
2. Literature review
2.1. Background
The literature review will describe and evaluate the development of the simulation on the impact of cycling on carbon emissions. The simulation tool involves many aspects, such as: traffic conditions, types of vehicles, driving behaviours and also traffic emissions. Therefore, it is important to evaluate the development step-by-step in order to produce a final simulation tool. Before performing the simulation tool, it is also important to review published articles in order to develop a better knowledge on how the traffic simulation process is conducted. Therefore, this literature review will give a thorough background understanding on the most widely used methods used to produce a reliable simulation tool. Section 2.2 compares and describes the trade-off between the simulation model and the analytical model. Section 2.5 describes the most widely used traffic models that are currently used - microscopic and macroscopic models. Section 2.4 describes the characteristics of the mixed vehicle traffic model. In this section, mixed vehicle is defined as the interaction between cars and bicycles and how bicycles significantly affect the performance of cars. Furthermore, Section 2.5 looks at the different traffic emissions models that are widely used nowadays, such as the average-speed based and the dynamic-based emissions model. Moreover, this part also gives a basic idea of what kinds of emissions pollutants are used in different published articles.
2.2. Simulation and Analytical model
2.2.1. Background of both models
Many published articles are based on research on transportation models in recent decades including transportation models such as freight, trains, vehicles, bicycles and other methods of public transport. However, implementing a traffic model for vehicles running on road networks requires the parameters of traffic flow operations, such as the reasons for traffic jams as well as the time and place when traffic breaks down [5]. Wilco [5] states that these kinds of traffic flow operations models have been widely researched and categorised into two dimensions: a simulation model and an analytical model. The simulation model provides a high level of detail on how the traffic process is performed, whereas the analytical model develops a mathematic equation to get an aggregate level of answers on traffic process.
2.2.2. Differences between both models
|
|
Simulation Model |
Analytical Model |
|
Model Complexity |
Unlimited |
Limited |
|
Run Time |
Long |
Short |
|
Data Requirements |
Large |
Small |
|
Model Development |
Predictable |
Unpredictable |
|
Flexibility |
High |
Low |
Figure 1. Trade-offs between the simulation and analytical model
By just comparing the models without considering the traffic conditions, each model has its own strengths and weaknesses, as shown in Table (1). The table concludes that the simulation model is more suitable for performing highly complex measures, whereas the analytical model only provides limited results of complexity with a lower level of detail. The computational running time is much higher in the simulation model due to the consequences of high complexity, whereas the analytical model has a relatively shorter running time. The high complexity of the simulation model requires a deeper knowledge of data in order to have a more accurate and reliable model. On the other hand, since the analytical model deals with an aggregate level of output and it requires less data collection. Model development refers to the model accuracy, where the simulation model is better in this category than the analytical model because the analytical model requires relatively less time and effort to develop. Lastly, the flexibility of the analytical model is less than the simulation model. This is because the analytical model is based on a mathematical equation and if the assumption of the equation changes then everything else must also be changed. Ekren [6] suggests that the analytical model evaluates the performance by using mathematical relationships between input and output, then these relationships are used to define an algorithm or to derive a formula and, as a result, these performance measures can be analysed. The simulation model simulates the sequence of many different events via a computer programme and is required to run many times to ensure the typical behaviour of the system is witnessed. To conclude, both models have their own trade-offs and no single model can address all problems. However, these are the general aspects of both models and they are not tailored to analyse traffic use. So, the next section is going to review the higher levels of both models in relation to traffic conditions.
2.3. Microscopic and macroscopic traffic model
2.3.1. Background
Having reviewed the characteristics and relationships between the simulation and analytical models, they give a general idea of what kinds of development processes this project needs. This section is going to look at how other published articles apply the simulation and analytical models with traffic conditions.
Traffic flow is unpredictable as traffic conditions evolve over time. A successful traffic simulation model requires the parameters of accuracy, representational and computational power [7]. Two types of approach are widely used nowadays; the microscopic and macroscopic approaches. Having discussed the simulation and analytical models in Section 2.2, the microscopic approach actually works like the simulation model, whereas the analytical model works like the macroscopic model. The advantage of having a microscopic model is that vehicle types are differentiated on the network such as passenger cars, buses and trucks [8]. Each bicycle and vehicle is treated as an identifiable entity [4]. By having a better accuracy of the model, the microscopic model provides details of driving behaviour such as road-following, car-following, traffic lights, overtaking lanes and many other traffic rules [8]. Therefore, vehicles react individually depending on how the users determine the simulation process. Again, Lopez-Neri et al. [9] describe the driving behaviours in the microscopic model from [8] as a “reactive, autonomous and social-multi-agent model”.
Nevertheless, Lopez-Neri et al. [9] argue that the microscopic model uses up too much computational resources due to the time-slicing method for updating agents and, as a result, the amount of time the agents interact in the road network is reduced. In addition, McCrea and Moutari [7] state that the microscopic model is not computationally competitive compared to the macroscopic model when it comes to model traffic in crowded and large road networks. Similarly, Nagel and Schleicher [10] also claim that the microscopic model runs entity individually to compute the possibility of the different routes that each vehicle may use on the network. As a result, most microscopic models used before 1994 were too slow to run on supercomputers. Later, they [10] overcame slow performance and increased the speed so that it was 1000 times faster by approaching a single-bit coding scheme. This paper is interested in the bicycle and vehicle (BV) model. The world’s first BV model from Faghri and Egyhaziova [4] not only includes driving behaviours rules, but also includes information about pedestrian behaviour, car parks and bus stops to further increase the complexity of the microscopic model. Therefore, modellers have been developing complex traffic models over the past seven decades [11].
Many macroscopic models have been developed in the past 50 years [12], and most of them are based on mathematical models and knowledge-based models [7]. Tyahi et al. [13] detail the main differences between the models. The microscopic model focuses on the car-following and lane-changing characteristics of individual vehicles, whereas the macroscopic model focuses on the aggregate behaviour of a collection of vehicles. Thus, Chakrobotry [14] states that the microscopic model defines the behaviour of dynamic traffic with the behaviour of individual vehicles driving in different situations, whereas the macroscopic model focuses on the inter-relationships among these three parameters: flow, speed and density. Similarly, McCrea and Moutari claim that these three parameters are the fundamental traffic parameters and are used in the macroscopic model [7]. In addition, Helbing [15] describes some of the advantages of using the macroscopic model. It allows users to control the probability of the model such as overtaking behaviour and it is also easier to calibrate compared with the microscopic model as less data is required. Ehlert and Rothkrantz [8] state that macroscopic models often derive from the fluid-dynamic model and every vehicle is treated the same. Therefore, the macroscopic model may not be very appropriate when it comes to developing a traffic model to deal with the interaction between vehicles and bicycles as both entities have different driving behaviours and cannot be treated identically. However, since the macroscopic traffic model uses a mathematical approach and does not need to collect a lot of information about traffic behaviour or conditions, the execution time would be significantly quicker than the microscopic traffic model. In addition, the output from the macroscopic model gives a total time or statistic based on fuel consumption rather than individual output from the microscopic model, but on the other hand it is less detailed than the microscopic model. Given that the microscopic and the macroscopic model both have unique features, some of the published articles have concentrated on research to combine both models to develop a hybrid traffic model called the mesoscopic model. This model provides the individual interactions between vehicles but their behaviour and interactions are based on aggregate level [5]. To conclude, the microscopic model is widely regarded as the simulation model, whereas the macroscopic model is regarded as the analytical model.
2.3.2. Computational implementation of traffic model
Both of the model has its different strength and weakness and therefore they have different approaches when it comes to computational implementation. The most common computational implementation of traffic simulation is called discrete-event system [43]. It means the system state stays the same unless a simulation event occurs, at which point that the system undergoes the state transition. A clock and a chronologically ordered event list manage the whole discrete-event system. When there is an event is running and once it is implemented, it changes the system state and also possibly triggers the schedule of other events. The first BV simulation model that mentioned previously also approach a discrete-event system, where it uses time-interval scanning to update the statement of the system. The computational implementation of the discrete-event system firstly to empty the road network, then it loads the vehicles and bicycles into the road network to perform the calculation [4]. To conclude, the whole mechanism of discrete-event simulation is as follow [43]:
Initial the simulation clock to an initial time and generates one or more initial events and schedule them.
Stop the simulation run if the event list is empty, otherwise find the most imminent event and unlink it from the event list.
Advance the simulation clock to the time of the most imminent event and excute it.
Loop back to step 2.
However, this is the general computational implementation for the discrete-event system.
2.4. Mixed vehicle traffic model
As previously discussed, the microscopic model has many different aspects on traffic behaviour, such as car-following, lane-changing and overtaking etc. However, only the car-following model has been widely researched, whereas the others have been investigated to a lesser extent. Therefore, this section will review the car-following model in greater depth. As previously discussed, the macroscopic traffic model focuses on the flow, speed and density of traffic behaviour. Therefore, the macroscopic traffic model has been widely used to develop congestion-related models. However, the congestion-related models are not directly related to this topic, since this report concentrates on the development of a model looking at the interaction between cars and bicycles.
2.4.1. Car-following model
Many papers state that the car-following model is a microscopic model [4, 8, 15, 16]. Ehlert and Rothkrantz [4] claim that the car-following model indicates that vehicles reduce speed to avoid crashing into the vehicle in front and the measurements of the gap and the speed between the vehicles are important. In addition, Jamison and McCartney [17] state that the car-following model simulates the motion of the individual vehicle on the road%u2028where the vehicle responds to the change of the vehicle ahead by accelerating or decelerating in a prescribed manner. Thus, Helbing [15] describes the car-following model in a similar manner to [4], but the paper further mentions lane-changing behaviour in multi-lane traffic. Similarly, Gunay [18] states that the key assumption of the car-following model is that vehicles are directly influenced by the front vehicle and overtaking is prohibited. In addition, Chakroborty [14] states that the car-following model has five different properties and these properties mainly outline that the car-following model should take human perception into account. For example, different humans have different reactions to distance and time when they try to accelerate or decelerate to avoid collisions. Moreover, Brackston et al. [19] define the car-following model as the process where different cars follow each other in the traffic network. However, the paper reviews many car-following models and several of them have not been thoroughly understood or validated.
