Predicting the Automated Vehicle Penetration using Mamdani Fuzzy Inference System for the city of Istanbul, Türkiye

Abstract

The advancement of automated vehicles is set to reshape transportation systems and urban development patterns, making it crucial to understand their future penetration. This study aims to predict the types and extent of automated vehicle (AV), shared automated vehicle (SAV), and shared automated electric vehicle (SAEV) penetration in Istanbul using a predictive fuzzy-based model. The Mamdani fuzzy inference system, incorporating five main criteria and nine sub-criteria, was applied to handle imprecise information from expert opinions and to simulate various planning scenarios from 2020 to 2060. The importance and motivation of this study lie in the need to anticipate and plan for the transformative impact of self-driving vehicles on urban transportation systems, ensuring that cities are prepared for the technological, infrastructural, and policy changes required for their integration. The research addresses the challenge of predicting how different types of self-driving vehicles will penetrate urban areas, helping to guide future planning and infrastructure development in Istanbul. The model results indicate that self-driving vehicles are expected to be concentrated in central business districts (CBDs) such as Kadıkoy, Beşiktaş, Sarıyer, and Üsküdar, driven by advanced transportation options and technological opportunities. Suburban residents, in contrast, are projected to prefer autonomous public transit over private AVs. Based on the predicted SAV and SAEV demands, there is a need to increase pick-up/drop-off points and charging stations, especially in districts like Çatalca and Silivri on the European side and Şile and Tuzla on the Asian side. Addressing parking challenges in the CBDs will also be necessary. The study’s findings offer valuable insights for transportation planners and decision-makers, highlighting the importance of enhancing electronic infrastructure, V2X communication, and reducing costs to promote autonomous vehicle adoption. Overall, this research contributes to future urban and transportation planning, helping predict driverless vehicle integration up to 2060 and guiding the design of essential infrastructure for Istanbul’s evolving mobility landscape.

Introduction

Development of leading technologies enables automated vehicles to reach the level of commercial introduction. Automated (or self-driving, autonomous, driverless) vehicle technology is one of the most prominent innovations and is supposed to play a major role in future mobility. The automated vehicles (AV) have a variety of advantages including time efficiency, accident prevention, increased mobility for those who are unable to drive and optimized traffic flow management which alleviates environmental pollution (Boesch, Ciari et al., 2016; Rahman and Thill, 2023). On the other hand, the potential concerns regarding AVs are thought to be high costs for usage, enhanced overall congestion and some potential risks such as cyber-attack risk.

Numerous auto manufacturers have been working on the completion of different level of automated vehicle technology, such as BMW and Ford which have planned to launch their self-driving vehicles in 2030 (Jungblut, Grube et al., 2024). The Victoria Transport Policy Institute reported that commercially availability of level 4-5 vehicles will increase in the 2020s, and be widely used by the 2040-2050s (Litman, 2020). AVs can be considered as a part of the “sharing economy”, which is the concept of sharing or renting vehicles instead of purchasing them, just like Uber, Lyft or shared taxis. Combination of sharing economy with the vehicle automation technology provides a new travel mode involving shared autonomated vehicles (SAVs). According to (Zhang, Guhathakurta et al., 2015), accessible mobility of the commuters will be enhanced by the new travel choice in a more affordable fashion when compared to conventional taxi. Another beneficial travel option is to use an electric vehicle (EV) which is mainly used in longer daily travel distances due to their relatively low energy and maintenance needs. A system of shared automated electric vehicles (SAEVs) could need large batteries and charging stations, yet may reduce the overall vehicle costs via lower energy and maintenance needs.

Automated vehicle penetration refers to the extent to which automated vehicles (AVs) have been adopted to the transportation system. Many recent researches have examined the potential effects of the widespread implementation of AVs on transportation systems and land use. (Van Arem, 2015) analyzed the physical transformation of cities pertaining to AVs by considering residential, employment, and recreational locations, and road, parking facilities. According to (Makahleh, Ferranti et al., 2024), the AVs will cause the removal of traffic signals and to reshape commercial or residential landscape due to decreasing land devoted to parking. (Thakur, Kinghorn et al., 2016) evaluated the automated vehicle development for Melbourne through developing land use and transport interaction model in which reduction value of travel time saving is utilized. Because AVs significantly increase the demand for road infrastructure the impact of AV on urban development depends heavily on the new technology that upcoming commercial models offer, including SAVs and SAEVs. According to Bakioglu et al. (2022), the shared driverless cars provide better mobility system with relatively lower cost. (Ratti and Biderman, 2017) reported that enhancing the usage of AVs in the era of shared technology may lead to either increase in driving since people could travel easily to central business districts (CBD) from suburbs or reduction of driving arising from providing more pedestrian-friendly CBD which, in turn will attract new residents.

Scenario planning was developed to address the future uncertainties, not to forecast, but consider the uncertainties as possible future outcomes (Ratti and Biderman, 2017). (Childress, Nichols et al., 2015) carried out four scenarios by modifying the capacity, perceived travel time, parking costs and operating costs of automated vehicles in which the reduction of vehicle mile traveled would cause an increase of operating costs from 0.15$ to 1.65 $. (Sunitiyoso, Wicaksono et al., 2023) used multi-stakeholder scenario planning to develop four future urban mobility scenarios for Jakarta, highlighting the need for strategies like vehicle electrification, MaaS, and behavior change to guide mobility over the next decade. (Kim, Yook et al., 2015) also used a scenario planning to analyze the effect of AVs on transportation and land use of the Seoul metropolitan city. They examined the development process of driverless vehicles with the reviews on proposed timeframe along with the optimistic and pessimistic views. The market share of AV for the years between 2020 and 2070 is estimated to increase from 3% to 100 %. By 2050, the assumption is 58 % of AVs that will be launched. In 2070, AVs production will be highly enhanced and completely used in urban patterns. (Heinrichs, 2016) shared that the areas including parking demand and organization, and the attractiveness of neighborhoods such as living, shopping and working places will be affected as a result of arising self-driving vehicles on existing transportation systems. (Litman, 2020) predicted the penetration of automated vehicles through the acquired experience with previous vehicle technologies, their benefits and costs. According to him, fully-automated vehicles would be manufactured for sale at the beginning of 2020s and widely used in the 2040–2050s.

According to (Rahman and Thill, 2023) launching different types of autonomated vehicles would have a major impact on urban mobility, sustainability and city designs. (Wang, He et al., 2024), reported that the efficient planning requires predicting future conditions and necessities accurately. For decision-makers, transportation and urban planners, how AVs will affect the future transits and knowledge of the distribution of AVs throughout the years are essential (Hao, Wang et al., 2023). The uncertainty of future conditions about transportation systems and vehicle technology along with the biased human estimations, leads to fuzziness in predicting the development of automated vehicle technology in urban formation. In traditional (crisp) set theory, the countless vague concepts that can easily be portrayed by human judgment, may not be reflected in a rational way. The classical methods may also not be capable of representing the uncertainties and complexities of future transportation system. Fuzzy system allows taking uncertain information into consideration in obtaining the model results (Zadeh, 1965). Fuzzy inference system (FIS) uses the fuzzy rules, which consist of verbal statements, to model the aspects of human judgments and predictions. According to (Kahraman, Cebeci et al., 2004), the fuzzy approach has intensive capability in modelling quantitative and qualitative problems involving vagueness and imprecision. Due to its accurate prediction and convenience for uncertain or approximate reasoning, fuzzy model has been widely carried out to engineering problems and evaluations processes such as sustainable transportation evaluation, selection of electric vehicle charging station, and supply chain management (Bakioglu, 2024; Miao, Zhao et al., 2024). To the best of authors’ knowledge, fuzzy approach has not been used for development of a model in predicting automated vehicle penetration in the literature.

Therefore, the objectives of this present study are to: i) predict the types of automated vehicle penetration involving automated vehicle (AV), shared automated vehicle (SAV) and shared automated electric vehicle (SAEV), ii) perform various planning scenarios to address the possible future outcomes regarding driverless vehicle pattern improvements for period between 2020 and 2060. Results are expected to be helpful in predicting future vehicle improvement and supporting transportation master planning and management.

Material and Methods

Mamdani fuzzy inference system (Mamdani-FIS), which was proposed by (Zadeh, 1965), is the process of a nonlinear mapping of a set consisting of input data sets into a set of output data based on fuzzy reasoning. Fuzzy logic allows decision making and evaluation with estimated values including uncertain information. Hence, the Mamdani approach can process verbal data (Altunkaynak, 2010; Şen, Zekai and Altunkaynak, 2006). Figure 1 depicts the basic structure of Mamdani-FIS showing three conceptual components: fuzzification, inference rules, and defuzzification.

Figure 1. The structure of Mamdani fuzzy inference system.

A fuzzy set is a collection of elements where each element has a membership value ranging from 0 to 1, representing the degree to which it belongs to the set. This allows for partial membership, making fuzzy sets suitable for handling uncertainty and imprecision in real-world problems.

Let X be a universe of discourse, x is the element of X. A picture fuzzy set, $\tilde{A}$ on X is defined as (Zadeh, 1965):

$\tilde{A}=\left\{⟨x,{\mu }_A\left(x\right)⟩\right|x\in X\}$ (1)

where ${\mu }_P\left(x\right)$ ϵ [0,1] denotes membership value. Complete non-membership is represented by 0, complete membership by 1, and values between 0 and 1 indicate partial membership in the fuzzy set $\tilde{A}$ .

Fuzzy numbers are a special type of fuzzy set that meet the conditions of normality and convexity, where

${sup}_{x\in X}{\mu }_A\left(x\right)=1$ (2)

${\mu }_A(k{x}_1+(1-k)x_2)\le min[{\mu }_A\left(x_1\right),{\mu }_A\left(x_2\right)]$ (3)

for all $x_1$ , $x_2$ ∈X and k ∈ [0,1]. Among different classes of fuzzy numbers, triangular and trapezoidal fuzzy numbers are commonly used in practical applications.

In the fuzzification process, crisp inputs are converted into linguistic fuzzy variables using membership functions (Şen, Zekâi and Altunkaynak, 2009). The membership function allows users to graphically represent a fuzzy set, with membership degrees varying between 0 and 1 (Zadeh, 1965). As fuzzy systems handle real-number inputs and outputs, the fuzzification process serves as an intermediary between the fuzzy inference engine and the environment. In this process, input values are treated as fuzzy singletons, and membership grades of all fuzzy propositions in the rule antecedents are evaluated. Fuzzification involves using the membership functions of linguistic variables to compute each term's degree of validity at specific points in the process. When a fuzzy rule activates, it fires to a certain degree based on the belief level in each antecedent. Commonly used membership functions include Triangular, Trapezoidal, Bell-shaped, and Gaussian functions, with their shapes chosen according to the application’s requirements.

According to (Kahraman, Cebeci et al., 2004), using triangular numbers is an efficient way to simulate decision making problems under the subjective and imprecise information. Therefore, in this present study, the triangular membership function is used for fuzzifying the crisp inputs. Triangular function is defined by a lower boundary a, an upper boundary b, and a value m, where a < m < b.

A (x) = $\{\begin{array}{c}\hfill 0,ifx\le a\hfill \\ \hfill \frac{x-a}{m-a},ifx\in [a,m]\hfill \\ \hfill \frac{b-x}{\begin{array}{c}\hfill b-m\hfill \end{array}},ifx\in [m,b]\hfill \\ \hfill 0.ifx\ge b.\hfill \end{array}$ (4)

The membership functions are divided into seven fuzzy sets labeled as 'Very Low' (VL), 'Low' (L), 'Medium Low' (ML), 'Medium' (M), 'Medium High' (MH), 'High' (H), and 'Very High' (VH). These linguistic terms provide a qualitative assessment of the variables and allow for a more nuanced representation of data, accommodating uncertainty and imprecision. Each linguistic value is converted into a corresponding triangular fuzzy number, as shown in Table 1, which captures the fuzzy set’s range and the degree of membership across the interval. Triangular membership functions were chosen due to their simplicity and ease of implementation, as they effectively represent fuzzy sets with three parameters (lower limit, peak, and upper limit). This seven-scale fuzzy linguistic variable system allows for a more precise modeling of input and output data, enabling a detailed analysis of the variables in the fuzzy inference system.

Table 1. Linguistic terms and fuzzy numbers

Linguistic Terms	Fuzzy Numbers
Very Low (VL)	(0, 0, 0.1)
Low (L)	(0, 0.1, 0.3)
Medium Low (ML)	(0.1, 0.3, 0.5)
Medium (M)	(0.3, 0.5, 0.7)
Medium High (MH)	(0.5, 0.7, 0..9)
High (H)	(0.7, 0.9, 1)
Very High (VH)	(0.9, 1, 1)

The inference rules play a crucial role in a fuzzy inference system (FIS) by establishing the relationship between input and output variables. These rules are typically expressed using linguistic terms in an ‘IF-THEN’ format, which is more intuitive than numerical representations (Şen, Zekâi and Altunkaynak, 2009). Each rule has two parts: the antecedent (premise) represented by 'IF' and the consequent (result) represented by 'THEN,' where the input subsets are usually combined using the logical 'and' conjunction, and the rules themselves are linked by the logical 'or' conjunction. The fuzzy “if-then” rules can be performed to show the relations between the input and output variables. For instance, ‘‘if x is low then y is high”, where the linguistic variable low and high are demonstrated by fuzzy numbers in the membership functions (Altunkaynak, 2014).

The degree to which each rule activates relies on how closely its antecedent fits the input; all rules with any truth in its premises will activate and add to the final fuzzy conclusion set. Because fewer rules are needed to specify the input-output connection and interpolation between various input states is made possible by this imperfect matching, the system operates more efficiently.

The utilization of 'IF-THEN' rules is essential, as they capture and embed human expertise and knowledge into the fuzzy framework, making it suitable for solving real-world problems. In this study, a set of fuzzy linguistic rules based on expert knowledge was defined to implement the proposed FIS model. The membership functions within these rules represent the linguistic terms, such as 'low' or 'high,' as fuzzy numbers, allowing for the flexible handling of uncertain and imprecise information. This approach ensures that the rules accurately reflect the preferences and knowledge of decision-makers, enabling the fuzzy inference system to model complex relationships effectively.

Finally, defuzzification is conducted in the last step to convert the fuzzy output into a crisp, numerical value that can be utilized for decision-making. This process is crucial, as it translates the fuzzy conclusions derived from the fuzzy inference engine into a single, precise output. The fuzzy inference process typically employs a 'min-max inference' method, using the minimum and maximum operators to determine the degree of membership for each rule's output. The resulting fuzzy output is a combination of these rules, reflecting all the influences with truth values greater than zero.

Various defuzzification methods are available, including the centroid (center of area, COA), bisector of area (BOA), mean of maximum (MOM), smallest of maximum (SOM), and largest of maximum (LOM) techniques. Among these, the centroid technique is the most widely used and involves calculating the center of gravity of the aggregated fuzzy set to obtain the crisp value. This technique ensures a balanced representation of all activated rules and provides a more accurate numerical output that represents the fuzzy inference result. It is defined as

Z* = $\frac{{\int }_{}^{}Z.\mu .\left(Z\right).dZ}{\mu .\left(Z\right).dZ}$ (5)

where Z*, Z and μ are crisp output, fuzzified value, and the assigned membership degree, respectively.

Study Area

Istanbul, Türkiye, which is located between coordinates 28°15'0.0" E to 29°30'0.0" E and 40°48'0.0"N to 41°33'0.0"N, is having traffic congestion problems every day. Istanbul is the largest metropolis in Europe with a population of over 15 million, divided into 39 districts, and having the total road network length of 25.000 km. As of 2024, the total number of vehicle is 4.644.743 and increasing by 30.000 every month. Existing transportation infrastructure includes 52 km bus rapid transit (BRT) carrying 900,000 people daily. The seaway is another option for transportation which serves 565.472 passengers per day. Railway infrastructure is also one of the most developing modes of transportation in Istanbul. By all means of transportation types, in total, 2.709.914 commuters are carried daily (TÜİK, 2024). In this research, 20 districts with an area of approximately 4.789 km2 and a population of almost 9 million, are studied (as shown in Figure 2).

Figure 2. Population Distribution for Assigned Regions.

The primary reason behind including 20 districts is to better represent the variety of counties having various population size and transportation facilities while keeping the computational complexity and execution time of calculation at a certain level. Intelligent transportation systems (ITS) have been developed in order to alleviate traffic congestion, emission, time spent in traffic as well as to increase mobility, traffic capacity, and road safety (Gokasar and Bakioglu, 2018). Developing intelligent transportation system facilitates the adoption of driverless car on roads. A number of ITS are readily installed in Istanbul such as Variable Message Sign (VMS), Lane Control System (LCS), Traffic Control Center Camera (TCC), and Electronic Inspection Camera (EIC). The locations of intelligent transportation systems in Istanbul are presented in Figure 3.

Figure 3. Locations of Intelligent Transportation Systems in Istanbul.

The automotive industry in Türkiye has been steadily evolving toward the production of autonomous vehicles (AVs), making significant strides through collaborative efforts between universities, independent companies, and government support. Türkiye's ambition to become a producer of future technology is evident in its ongoing projects and partnerships, despite being behind global leaders like Germany and the United States in the autonomous vehicle sector.

Key initiatives, such as the Türkiye Connected and Autonomous Vehicle Cluster (TCAV), were established to foster collaboration among 62 organizations, including TOSB (Automotive Supplier Industry Specialized Organized Industrial Zone) Innovation Center and ITU OTAM (Automotive Technologies Research Development Center). Companies like ADASTEC have moved forward with live traffic tests for their autonomous vehicles, utilizing LIDAR technology for infrastructure mapping. Eatron, operating out of both Istanbul and the UK, focuses on AI-based intelligent software for mass production electric and autonomous vehicles.

Furthermore, the partnership between Otokar and University has led to the development of an advanced autonomous bus system, marking a significant step in autonomous vehicle technology in Türkiye. Meanwhile, TOGG (Türkiye's Automobile Initiative Group) has announced the launch of their Level 3 autonomous electric vehicle, showcasing Türkiye's progress in this field.

The first autonomous vehicle test drives in Istanbul began in 2021 with Eatron, indicating the city's growing role as a center for AV testing and development. Istanbul's ITU ARI Teknokent serves as a hub for ongoing autonomous vehicle research and development, and these activities are gradually shaping the technological landscape for AVs in Türkiye's largest city.

According to data from the Turkish Statistical Institute (TÜİK), as of 2024, automobiles account for 52.2% of all registered vehicles in Turkey. The remaining vehicle types were provided Figure 4. This distribution highlights the significant presence of automobiles in the Turkish vehicle landscape (TÜİK, 2024).

Figure 4. Vehicle distribution in Turkey.

A notable trend is the increasing adoption of electric vehicles (EVs). During the period from January to August 2024, there were 683,918 newly registered automobiles, categorized by fuel type. Among these, petrol vehicles made up 64.1%, followed by hybrids at 13.9%, diesel vehicles at 12.6%, and electric vehicles at 8.3%. Additionally, LPG vehicles constituted 1.1% of the new registrations (TÜİK, 2024).

Moreover, the majority of new vehicles are equipped with autonomous features, with at least Level 1 autonomy available. Some models even offer Level 2 autonomy options, such as adaptive cruise control and lane-keeping assist, indicating a shift towards more advanced driving technologies in the automotive market.

Although Türkiye's automotive industry is advancing rapidly, the average age of automobiles is 14.5 years, indicating that many vehicles on the road lack modern autonomous technology. This presents a significant challenge but also highlights the potential for AVs and EVs to transform the market as newer models with advanced autonomous features become more prevalent.

Istanbul, with its growing technological capabilities and collaborative efforts, is poised to be a significant player in the autonomous vehicle industry. The city’s AV penetration can play a critical role in the transformation of transportation systems at both local and national levels, paving the way for broader integration of AVs and EVs in the urban landscape.

Figure 5 illustrates the distribution of various vehicle types across Istanbul's selected districts, including private vehicles, commercial vehicles, motorcycles, bicycles, and electric scooters.

Figure 5. Distribution of vehicle types across Istanbul's selected districts.

Private vehicles dominate in all districts, with the highest numbers in Kadıköy (597), Pendik (650), and Üsküdar (538), indicating high ownership that may reflect affluence or dependence on personal transportation. Commercial vehicles are less common, with Üsküdar leading at just 11, while other districts report only 2, suggesting that commercial activities are concentrated in specific areas. Motorcycle usage is notable in Kadıköy (58) and Üsküdar (33), indicating that local traffic conditions may favor this mode of transport. Bicycles are also prominent in Kadıköy (40) and Üsküdar (50), likely due to better cycling infrastructure and a cultural preference for sustainable transportation. Çekmeköy (35) shows a significant number as well. Electric scooters, while still low overall, are most used in Kadıköy (17), Çekmeköy (14), and Büyükçekmece (12), reflecting their emerging viability influenced by environmental awareness.

The visualization reveals distinct patterns of vehicle usage across the districts of Istanbul, reflecting diverse transportation preferences shaped by local infrastructure. At the same time, Istanbul is emerging as a significant hub for autonomous vehicle (AV) technology in Türkiye. The commencement of autonomous vehicle test drives in the city in 2021 underscores Istanbul's commitment to AV innovation. Additionally, the increasing registration of electric vehicles, along with the growing prevalence of new models equipped with at least Level 1 autonomy, highlights the transformative potential of AVs and EVs in reshaping urban transportation. With its advancing technological capabilities, Istanbul is well-positioned to integrate AVs into its transportation framework, thereby significantly influencing both local and national systems.

The public sentiment toward autonomous vehicles (AVs) in Istanbul is marked by a significant degree of unfamiliarity, reflecting a general hesitance to embrace this emerging technology. Research conducted by (Bakioglu, Salehin et al., 2022) reveals that a substantial portion of the population in Istanbul, approximately 35%, identifies as moderately unfamiliar with self-driving vehicles. This lack of familiarity is further emphasized by the fact that 32.7% of respondents consider themselves very unfamiliar with AVs. Only a small segment of the population, approximately 3.7%, feels very familiar with these vehicles, indicating a considerable gap in awareness and understanding of AV technology. Figure 6 shows the distribution of autonomous vehicle familiarity in Istanbul.

Figure 6. Distribution of autonomous vehicle familiarity.

Further analysis indicates that gender plays a notable role in public sentiment towards AVs. Females tend to have more negative opinions of AVs and are less willing to pay a premium for such technology compared to males. Conversely, individuals in Istanbul who express a willingness to invest more in acquiring new vehicles tend to have a more positive outlook on autonomous vehicles. This suggests that economic factors and familiarity with technology significantly influence public perception.

The acceptance of different levels of automation also varies. Partial and full automation levels of AVs are relatively well-accepted by Istanbul residents, particularly for private vehicles (PV) and shared vehicles (SV). The familiarity with partial automation, which is already available in the global market, leads to more positive acceptance. Additionally, the potential benefits of full automation, such as reducing driving burdens and human errors, contribute to positive perceptions. However, conditional automation, which includes some limitations like restricted access to certain highways, is not as well received, likely due to a lack of familiarity and perceived benefits.

Data Collection

Automated vehicle technology is a futuristic system involving technological landscape, environmental, economic, transportation, risk, and social factors. The criteria for determining the types of automated vehicle penetration include 5 main criteria together with 9 sub-criteria which are indicated in Table 2. These criteria are obtained mainly from extensive literature surveys and experts’ knowledge.

The innovational factors such as digital infrastructure, explainable artificial intelligence (XAI), are physical infrastructures changes that help the adoption of AVs on roads. Road markings, crossings and junctions, and charging stations have to be perceived by self-driving cars in order to commute autonomously. With the advancement of artificial intelligence (AI) techniques, automated vehicle can sense their surrounding and make reasonable decisions based on the digital infrastructure’s state, through the help of real-time and historical knowledge-based data (Bakioglu, 2025). The fifth-generation (5G) technology is also essential for connecting vehicles with other nearby vehicles, and urban infrastructure. The 5G technology can specifically transmit all required data, such as real-time data regarding current state of roadway (Hakak, Gadekallu et al., 2023).

Ecological factors are associated with environmental pollution and energy usage. Driverless cars reduce the total energy consumption and air pollution by 4% to 25% since trailing distance between the vehicles are so close, less air resistance is experienced (Aini, Shen et al., 2023). It could be expected that the shared automated electric vehicle (SAEV) will provide more reduction of energy usage and air pollution than automated vehicle (AV) and shared automated vehicle (SAV) (Jungblut, Grube et al., 2024).

The economic factors including cost of self-driving cars, are another criterion for evaluation of those of cars. SAVs and SAEVs are developed to serve as public transportation vehicles; therefore, usage cost of them is likely to be less than that of AVs. According to (Boesch, Ciari et al., 2016), the technology can reduce some of the costs associated with driving, such as gasoline and insurance costs.

Transportation criteria includes vehicle kilometer travelled (VKT), vehicle miles travelled (VMT) and parking demands. Studies showed that the usage of private AVs increase vehicle kilometer travelled and reduce the public transit modes. (Kim, Yook et al., 2015) presented that, 24% increase in VMT causes 50% and 100% decrease in value of time (VOT) and parking costs, respectively. On the other hand, SAVs is found to decrease the overall number of vehicles (about 25-60%) and thus, parking spaces (Childress, Nichols et al., 2015).

Safety factors could vary among the types of self-driving cars. The importance of perceived risk can be thought as a key component of technology acceptance. The perceived risk is linked to a decision which is whether to use an automated device or not. The technology acceptance and perception of risks for the future vehicle types may differ among the districts. The possibility of road accidents with AVs might be the primary source of hesitation for the adoption of those vehicles. (Bakioglu and Atahan, 2021) stated that the potential for vehicle collision has adverse impact on the intention to use automated vehicles.

Table 2. The criteria for determining the types of autonomous vehicle penetration.

Criteria	Sub-Criteria	Definition
Innovational Factors	Digital Infrastructure (DI)	Interaction of vehicle with other vehicles, pedestrians, infrastructure and the network, as well as to navigate autonomously.
	Explainable Artificial Intelligence (XAI)
	The fifth-generation technology (5G)
Ecological Factors	Air Pollution Impact (AP)	Reflection of environmental pollution and energy usage.
	Energy Efficiency (EE)
Economic Factors	Aspect of Cost (AC)	Relevance to cost of self-driving cars.
Transportation Factors	Vehicle Kilometer Travelled (VKT)	Different types of driverless car may be affected by traffic related factors in different way.
	Impact on Parking Spaces Necessity (PS)
Safety Factors	Aspect of Accident Risk (AR)	Perception of risks for the future vehicle types may differ by districts.

The calculation of vehicle penetration for the assigned districts involves a comprehensive process that integrates multiple input factors using a fuzzy logic-based model. This process is executed in several stages, each incorporating different elements of fuzzy logic theory to ensure accurate predictions. Below, a step-by-step analysis is provided, with examples, showing how inputs are used to calculate vehicle penetration for AVs, SAVs, and SAEVs.

Step 1, defining the inputs: The inputs for calculating vehicle penetration in different districts include key factors influencing the adoption and integration of autonomous vehicles (AVs). These inputs encompass aspects such as innovational, ecological, economic, transportation, and safety factors.

Step 2, fuzzification: The fuzzification process converts crisp input values into fuzzy values representing degrees of membership in fuzzy sets. For this study, a triangular membership function was used, and inputs were divided into seven linguistic terms. For example, the 'Digital Infrastructure' input for Kadıköy might be rated as 'Medium' (M), whereas for Şile, it could be 'Low' (L).

Step 3, creating the knowledge base and fuzzy rules: The knowledge base contains fuzzy rules that guide the decision-making process. Fuzzy rules are expressed as “If–Then” statements. For instance:

•    For Kadıköy, if DI, XAI, and 5G are all 'Medium' (M), then the expected penetration for AV is 'High' (H), while SAV and SAEV are 'Medium High' (MH).
•    For Şile, if DI, XAI, and 5G are all 'Medium' (M), then AV is 'Medium' (M), SAV and SAEV are 'Medium Low' (ML).

These rules are created by experts based on the district's characteristics, ensuring that each district’s unique properties are accounted for.

Step 4, defuzzification: The defuzzifier converts the fuzzy output values into crisp numerical values using the centroid method. This step provides the final vehicle penetration levels for AV, SAV, and SAEV for each district, translating fuzzy outputs into concrete predictions.

Results

The types of self-driving car penetration levels are depicted as 3-D column chart in Figure 7 and Figure 8 for European and Asian sides of city, respectively. The full self-driving vehicles have not been currently launched. It is considered that buying opportunities were available by 2030, accordingly, the effects of aforementioned criteria on the development of driverless cars type will start to be revealed.

In the Asian side, Kadıkoy (KD) and Uskudar (US) have the largest number (i.e. around 0.37) and Sile (SL) has the smallest number (i.e. around 0.25) of AVs, SAVs, and SAEVs due to the fact that the population and its growth rate in Kadıkoy and Uskudar are higher than those in Sile. (Abbreviations of KD, US, BE, UM, SL, CK, TZ, PN, ML, KR represent the districts of Kadıkoy, Uskudar, Beykoz, Umraniye, Sile, Cekmekoy, Tuzla, Pendik, Maltepe and Kartal, respectively.) In the European side, autonomous vehicle usage is at the highest level in Besiktas (BS) which has the highest population and life quality (i.e. 0.38); however, Arnavutkoy (AR), which owns the worst socio-economic indicators, has the least self- driving cars numbers (i.e. around 0.25). (Symbols of BS, SR, ZY, SS, CT, SL, BC, AR, BK, EY state the districts of Beşiktaş, Sarıyer, Zeytinburnu, Şişli, Çatalca, Silivri, B.Çekmece, Arnavutköy, Bakırköy, Eyüp, respectively). The central business districts (CBD) have much more internet accessibility and parking lot opportunities, and have more intelligent transportation systems (Figure 3) accommodate higher number of self-driving vehicles compared to suburban districts. Kadikoy, Besiktas, Sariyer, and Uskudar provide relatively more free internet access and parking lots, and have relatively more advanced transportation infrastructures which may facilitate the usage of driverless vehicles. However, these districts have higher number of vehicles and related accidents; therefore, they have higher greenhouse gas emission. Sile and Arnavutkoy which are far away from the CBD are having internet connection problem, which might reduce the adoption of self-driving vehicles. Suburban roads, also do not own intelligent systems, and the driverless cars need intelligent infrastructure as discussed previously.

Figure 7. Types of self-driving vehicles penetration in the Asian side of the city.

Figure 8. Types of self-driving vehicles penetration in the European side of the city.

Figure 9 shows the spatial distribution of AVs, SAVs, and SAEVs in Asian and European sides, respectively. It can easily be seen that self-driving vehicles are concentrated in CBD, such as Kadıkoy, Besiktas, Sarıyer and Uskudar. Numerous business and educational centers, and employment areas are located at CBDs which led to larger values of AVs. The residents who live in the suburbs mostly use variety of transportation modes for reaching to central business district. The suburban roads have also a variety of intelligent transportation system which might facilitate the self-driving vehicle movements. Car pooling and public transportation options are specifically preferred among mode of transport; therefore, suburban districts such as Pendik (PN), Tuzla (TZ) and Catalca (CT) own the larger portion of SAV and SAEV than AV.

Figure 9. Spatial distribution of driverless car penetration in European and Asian continents, respectively.

The typical transportation scenario planning includes urban development and advancement of technology related to transportation patterns for the horizon of 20 to 50 years. In this present study, six types of scenarios are applied to examine the evaluation of driverless vehicle development for assigned decades (in Table 3). Different types of automated vehicles penetration are predicted spanning the years 2025 to 2060. While the inclusion of the 2020 prediction serves to facilitate a comparative analysis between existing situation and the subsequent model results.

It is imperative to acknowledge that this paper is composed in 2024. As such, the predictions for 2025 and beyond are prospective estimations based on the scenarios presented in Table 3. Different decades would affect the self-driving vehicle penetration dynamics as shown in Figure 10. The values range between 0 and 1 in the Figure 10, indicating the estimated proportion of automated vehicles within each district or region. For instance, a value of 0.5 shows that automated vehicles make up 50% of the total vehicles in that district. Similarly, a value of 0.25 would represent a 25% penetration rate.

As the time passes by, the frequency of driverless car employment increase showing at the same time the varying trends among different districts. During the time period between 2020 to 2060, the autonomous vehicle penetration of downtown districts of both continents such as Kadıkoy (KD), Umraniye (UM), Besiktas (BS), Sisli (SS) would increase 77%, 64%, 72%, and 65%, respectively. On the other hand, the autonomous public vehicles such as SAV and SAEV facilitates the increase of the automated vehicle penetration more than that of AVs in the suburban districts compared to the previous year. As the years goes by, therefore, as foreseen, suburban areas such as Pendik, Cekmekoy, and Catalca SAVs and SAEVs employment increase by 90%.

Table 3. Scenario lists

Scenario	Description	Decade
S1	Level 4-5 vehicles become commercially available	2020s
S2	Level 5 vehicles are available having moderate price premium	2025s
S3	Level 5 vehicles will be sold much and charging stations for electric vehicles will vary	2030s
S4	Automated car sharing or taxi will be widely used	2035s
S5	Most vehicle are capable of autonomous driving	2040s
S6	Mandated autonomous vehicles	2060s

The increasing rates of driverless vehicle types may differ for each city, illustrating that region's mode preference. Moving away from the CBD may cause an increase in automated public transit usage. As illustrated in Figure 10, Catalca and Silivri in European side and, Sile and Tuzla in Asian side, which are far away from the downtown have the higher rates of SAEV and SAV usage than that of AV. According to Figure 10, the SAEV usage would increase 15% compared to SAV and AV usage after the year 2040 and in 2020, 2025, and 2030 the automated vehicle usage would be superior to rest of automated vehicle types. However, in 2035, shared automated vehicle usage would increase compared to the other vehicle types. Automated vehicle usage in European side is higher than that in Asian side due to more condensed employment areas and workplaces located in the European side.

As obviously illustrated in Figure 11, automated vehicle, shared automated vehicle, and shared automated electric vehicle employment in Istanbul differs across six scenarios. For Scenario-1 and Scenario-2, automated vehicles would be introduced and be commercially available. Therefore, automated vehicles are more densely distributed than shared and electric ones. For Scenario-3 and Scenario-4, automated public transits, particularly shared automated vehicle, show an increase especially at the suburbs. It can be illustrated that the districts with less public transportation options would have the higher SAVs usage than other modes. With the increase in charging stations for electric vehicles the SAEVs would be used a little bit more than those predicted in years between 2020 and 2025. As regards to Scenario-5 and Scenario-6, most of the vehicles shift from manual to automatic gear and the driverless vehicle is dominantly employed for travel. As can be deduced from the related scenarios, there exists a tendency for people to use SAVs than other types of vehicles. The overall cost of an automated vehicle is higher than that for shared and shared electric vehicles; therefore, wealthy people may rather choose to use the automated vehicle. All scenarios clearly indicate that people who live in central business district would prefer automated vehicles to other types of vehicles. On the contrary, people who live in the suburbs would mostly prefer shared automated electric vehicle to reach to CBD.

An important element of the policy improvement and implementation process is to estimate the future condition of the city (Kim, Yook et al., 2015). This information can inform planning scenarios, including the penetration of self- driving vehicles. Results obtained from Scenario-1 can be compared with actual state of affairs in 2020. In the first scenario of this study, a prediction was made that Level 4-5 automated vehicles would become commercially available by 2020. However, as the industry stands in the present, this prediction has not materialized as predicted. While significant progress has been made in the testing of autonomous vehicle technology, the widespread availability of Level 4 automated vehicles for consumer usage remains limited. Similarly, Level 5 automated vehicles, which represent full automation under any conditions, have not yet reached the stage of widespread commercial deployment. Thus, the initial prediction regarding the availability of Level 4-5 automated vehicles in 2020 might be optimistic.

The year-based scenarios indicate that car sharing membership, in particular shared automated electric vehicles show an increase from 2025 to 2060 on yearly basis. Growing shared electric vehicle employment has important impacts on greenhouse gas emissions and energy consumption. The plug-in SAEV adoption and SAV usage will help increase the air- quality standards, traffic capacity utilization, mobility and energy efficiency, and decrease the rate of carbon-emissions, fuel consumption and overall traffic congestion. There could also be some drawbacks for electric vehicles, especially at the first stage of the advent of shared automated electric vehicles. Bartlett, (2012), stated that users might have ‘‘range anxiety”, which is the fear of insufficient range of the vehicle in reaching its destination, when they use electric-cars. Figure 12 shows the locations of charging stations for the electric vehicles. From the years 2025 through 2030, Kadıkoy (KD), Uskudar (US), Kartal (KR), Maltepe (ML), Besiktas (BS), Sarıyer (SR), Zeytinburnu (ZY) and Buyukcekmece (BC) districts have the higher number of shared automated electric vehicles when compared to the other automated vehicle types. With the increased amount of charging stations for electric vehicles by 2030s, passengers may overcome the “range anxiety” and to start using shared automated electric vehicles. As can be inferred from Figure 11, SAEVs will be the most preferred transportation option after the year 2030. Thus, the rate of SAEV usage may increase after the charging stations are enhanced for the accommodation of SAEVs. It is to note that, shared automated vehicles pick up the travelers from stations and thus, different forms of infrastructure may be required which will have to be the taken into consideration by transportation planners. The new pick up points will eventually have to be designed in districts based on the SAV requirements and pick up point locations could be inferred from this study. The number of pick up and drop off points, and charging stations should be increased particularly in Catalca and Silivri for European side, and Sile and Tuzla for Asian side.

Figure 10. Variation in types of self-driving cars penetration by the decades [(a1-a2) AVs, (b1-b2) SAVs, (c1-c2) SAEVs, and numbers of “1”, “2” represent Asian and European continents, respectively]

Figure 11. Self-driving cars penetration under different scenarios [(a1-a3) 2020, (b1-b3) 2025, (c1-c3) 2030, (d1-d3) 2035, (e1-e3) 2040, (f1-f3) 2060, and the number of “1”, “2”, “3” represent AV, SAV, and SAEV penetration].

Dicussion

The introduction of automated vehicles (AVs) would significantly enhance the mobility of individuals, including children, the elderly, and disabled people. As observed by Makahleh, Ferranti et al. (2024), AVs have the potential to improve road safety, reduce traffic congestion, air pollution, and fuel consumption, while simultaneously enhancing accessibility for those with mobility challenges. Consequently, this aligns with our findings, as shared automated vehicles (SAVs) and shared automated electric vehicles (SAEVs) will likely be more prevalent in suburban areas, where there is a greater need for enhanced mobility solutions. The study also indicates that the increasing rates of driverless vehicles will demand new parking strategies and pricing systems that will effectively manage congestion and avoid further traffic issues, especially in central business districts (CBDs). (Heinrichs, 2016) suggested that AVs will influence the organization of parking demand and impact neighborhood attractiveness for living, shopping, and working, a trend likely to be more pronounced in downtown areas with the highest AV adoption.

Moreover, the findings of this study indicate that shared automated vehicles may lead to urban sprawl as individuals might decide to live further from city centers, as reported by (Duarte and Ratti, 2018). However, (Ewing, Hamidi et al., 2016) highlighted that the potential increase in fatal car crashes associated with urban sprawl might be mitigated by AVs due to their enhanced safety features. Thus, AVs could play a pivotal role in reshaping urban planning and reducing the negative impacts typically associated with urban sprawl.

Considering the impact of shared automated electric vehicles (SAEVs) on sustainability, (Nemoto, Issaoui et al., 2021) emphasized that the deployment of SAEVs in public transportation contributes to multiple sustainability goals, including environmental, social, and economic aspects. This reinforces our study's implication that the increasing use of SAEVs will reduce greenhouse gas emissions, improve energy efficiency, and enhance traffic flow in urban areas. For example, our study predicts a 15% increase in SAEV usage after 2040, which will positively impact air quality and reduce carbon emissions.

Furthermore, as electric charging stations become more widespread, the anxiety associated with the limited range of electric vehicles is likely to diminish, encouraging more people to adopt SAEVs. Bakioglu (2024) further emphasizes the importance of integrating sustainable transportation strategies within different contexts, such as campuses, which mirror broader urban environments. This insight is valuable for urban planners and policymakers when considering the integration of AVs, SAVs, and SAEVs, as similar sustainable strategies can be adapted for city-wide implementation to promote a more environmentally friendly and efficient transportation system.

The findings of this research suggest that decision-makers should develop policies and strategies that support the infrastructure needs of AVs, SAVs, and SAEVs, such as increasing the number of charging stations, especially in suburban areas. Additionally, congestion pricing should be considered for private autonomous vehicles in CBDs to regulate demand and minimize traffic congestion. Moreover, enhancing public transit options with autonomous buses, bus rapid transit (BRT), and automated minibuses will further increase mobility and reduce traffic congestion in the future. These insights offer a comprehensive understanding for policymakers, stakeholders, and urban planners, providing a foundation for developing long-term strategies for the integration of autonomous vehicles into urban transport systems while ensuring sustainability and efficiency.

Conclusion

This present study has developed a fuzzy based model to evaluate the self-driving vehicle penetration under various scenarios for the city of Istanbul, Turkey. The fuzzy model provides accurate prediction results under vague information, accordingly, facilitates capturing the knowledge-based experts’ judgments for future conditions about transportation systems and future vehicle technology. In addition, fuzzy model could enable researchers to use both quantitative and qualitative data simultaneously during assessment process. Implementation of the developed fuzzy model to the city of Istanbul yielded that, in the Asian side, Kadıkoy (KD) and Uskudar (US) have the largest number of driverless vehicles with the quantity of 0.37, while Sile (SL) owns the smallest number with the quantity of 0.25. As regards the European side, Besiktas (BS) has the highest level in automated vehicle penetration in with the number of 0.38, while Arnavutkoy (AR) with the quantity of 0.25 has the least level for self-driving cars. According to the analysis carried out in this present study, people who live in central business districts would like to prefer mostly automated vehicles to the other options. The urban roads having intelligent transportation systems, various employment and living areas may significantly affect the decision of whether adopting self-driving vehicles or not.

This study reveals potential implications on urban penetration improvement of transportation system with driverless vehicles and assessing the possible political and economic decisions. Based on the existing scenarios and their ideas involving integrating types of automated vehicles into transport systems, different improvements can be envisaged. Scenario analyses suggest that increasing automated driving capabilities and charging stations may promote the driverless vehicle usage in the city of Istanbul. Enhancing the electronic infrastructure for automated vehicles, V2X communication among vehicles and decreasing the cost of the automated vehicles are efficients way to improve different types of automated vehicle usage. As a result of scenarios, automated vehicles are mostly condensed in central business districts such as Kadıkoy (KD), Besiktas (BS), Sarıyer (SR) and Uskudar (US) while suburban people widely use automated public transits. The automated public transportation options may provide reduction in air pollution, more energy efficiency and capacity gain in traffic. Therefore, central business districts need to pay attention to both traffic congestion and air pollution due to higher automated vehicle usage projection. The results of this study present an essential contribution that could be useful for transportation planners and decision makers to predict the future transportation development. Planners and policy makers may benefit from this research for future prediction of most prominent districts in the city of Istanbul as far as year 2060. They could observe the future condition of the regions by reviewing this present study and use the obtained results to make decisions about designing new required areas such as charging station, pick up point, and parking allocation.

The prediction of self-driving vehicle usage can be more robust with some improvements. The number of evaluation criteria might be increased in order to obtain more accurate results. In addition, fuzzy model can be improved by incorporating land demand components which could reveal the required supply, such as potential residential lots, transport systems, catchment areas, for associated areas. The spatial distribution of charging stations for electric vehicles and parking lots, which can be obtained from the existing model, also play an important role in improvement of automated vehicle types.

Author Contributions

All authors have read and agreed to the published version of the manuscript. Conceptualization, G.B.; methodology, A.A. and G.B.; software, G.B.; investigation, G.B.; data curation, G.B.; writing—original draft preparation, A.A. and G.B.; writing—review and editing, A.A. and G.B.; supervision, A.A. All authors have read and agreed to the published version of the manuscript.

Ethics Declaration

The authors declare that they have no conflicts of interest regarding the publication of the paper.

Funding

No funding was received to assist with the preparation of this manuscript.

Acknowledgments

This research did not receive any grant from funding agencies in the public, commercial, or governmental sectors.

References

•  Aini,  N.,  Shen,  Z., et al. (2023). "Evaluation of the Shape of Tree Crowns to Protect Air Quality on the Roadside from the CO2 Dispersion Produced by Transportation". International Review for Spatial Planning and Sustainable Development, 11(2), 26-41. doi: https://doi.org/10.14246/irspsd.11.2_26.
•  Altunkaynak,  A. (2010). "A Predictive Model for Well Loss Using Fuzzy Logic Approach". Hydrological Processes, 24(17), 2400-2404. doi: https://doi.org/10.1002/hyp.7642.
•  Altunkaynak,  A. (2014). "Predicting Water Level Fluctuations in Lake Michigan-Huron Using Wavelet-Expert System Methods". Water Resources Management, 28, 2293-2314. doi: https://doi.org/10.1007/s11269-014-0616-0.
•  Bakioglu,  G. (2024). "Selection of Sustainable Transportation Strategies for Campuses Using Hybrid Decision-Making Approach under Picture Fuzzy Sets". Technological Forecasting and Social Change, 206, 123567. doi: https://doi.org/10.1016/j.techfore.2024.123567.
•  Bakioglu,  G. and  Atahan , A. O. (2021). "AHP Integrated TOPSIS and VIKOR Methods with Pythagorean Fuzzy Sets to Prioritize Risks in Self-Driving Vehicles". Applied Soft Computing, 99, 106948. doi: https://doi.org/10.1016/j.asoc.2020.106948.
•  Bakioglu,  G.,  Salehin , M. F., et al. (2022). "Examination of the Role of Safety Concerns from Autonomous Vehicle Ownership Choice: Results of a Stated Choice Experiment in Istanbul, Turkey". Transportation Letters, 14(10), 1172-1183. doi: https://doi.org/10.1080/19427867.2021.2007679.
•  Boesch , P. M.,  Ciari,  F., et al. (2016). "Autonomous Vehicle Fleet Sizes Required to Serve Different Levels of Demand". Transportation Research Record, 2542(1), 111-119. doi: https://doi.org/10.3141/2542-13.
•  Childress,  S.,  Nichols,  B., et al. (2015). "Using an Activity-Based Model to Explore the Potential Impacts of Automated Vehicles". Transportation Research Record, 2493(1), 99-106. doi: http://dx.doi.org/10.3141/2493-11.
•  Duarte,  F. and  Ratti,  C. (2018). "The Impact of Autonomous Vehicles on Cities: A Review". Journal of Urban Technology, 25(4), 3-18. doi: https://doi.org/10.1080/10630732.2018.1493883.
•  Ewing,  R.,  Hamidi,  S., et al. (2016). "Urban Sprawl as a Risk Factor in Motor Vehicle Crashes". Urban Studies, 53(2), 247-266. doi: https://doi.org/10.1177/0042098014562331.
•  Gokasar,  I. and  Bakioglu,  G. (2018). "Modeling the Effects of Real Time Traffic Information on Travel Behavior: A Case Study of Istanbul Technical University". Transportation Research Part F: Traffic Psychology and Behaviour, 58, 881-892. doi: https://doi.org/10.1016/j.trf.2018.07.013.
•  Hakak,  S.,  Gadekallu , T. R., et al. (2023). "Autonomous Vehicles in 5G and Beyond: A Survey". Vehicular Communications, 39, 100551. doi: https://doi.org/10.1016/j.vehcom.2022.100551.
•  Hao,  M.,  Wang,  X., et al. (2023). "The Elderly Acceptance of Autonomous Vehicle Services in Beijing, China". International Review for Spatial Planning and Sustainable Development, 11(1), 64-84. doi: http://dx.doi.org/10.14246/irspsd.11.1_64.
•  Heinrichs,  D. (2016). "Autonomous Driving and Urban Land Use". Autonomous Driving: Technical, Legal and Social Aspects, 213-231.
•  Jungblut,  E.,  Grube,  T., et al. (2024). "Automated and Connected Driving: State-of-the-Art and Implications for Future Scenario Analysis". arXiv preprint arXiv:2408.11864.
•  Kahraman,  C.,  Cebeci,  U., et al. (2004). "Multi-Attribute Comparison of Catering Service Companies Using Fuzzy Ahp: The Case of Turkey". International Journal of Production Economics, 87(2), 171-184. doi: https://doi.org/10.1016/S0925-5273(03)00099-9.
•  Kim,  K.-H.,  Yook,  D.-H., et al. (2015). "An Analysis of Expected Effects of the Autonomous Vehicles on Transport and Land Use in Korea". New York University: New York, NY, USA.
•  Litman,  T. (2020). "Autonomous Vehicle Implementation Predictions: Implications for Transport Planning".
•  Makahleh , H. Y.,  Ferranti , E. J. S., et al. (2024). "Assessing the Role of Autonomous Vehicles in Urban Areas: A Systematic Review of Literature". Future Transportation, 4(2), 321-348. doi: https://doi.org/10.3390/futuretransp4020017.
•  Miao,  L.,  Zhao,  P., et al. (2024). "Factor Analysis of Ev Purchase Behavior in Medium-Sized Cities in China: Case Study of Luoyang City". International Review for Spatial Planning and Sustainable Development, 12(2), 79-98. doi: https://doi.org/10.14246/irspsd.12.2_79.
•  Nemoto , E. H.,  Issaoui,  R., et al. (2021). "How to Measure the Impacts of Shared Automated Electric Vehicles on Urban Mobility". Transportation Research Part D: Transport and Environment, 93, 102766. doi: https://doi.org/10.1016/j.trd.2021.102766.
•  Rahman , M. M. and  Thill,  J.-C. (2023). "Impacts of Connected and Autonomous Vehicles on Urban Transportation and Environment: A Comprehensive Review". Sustainable Cities and Society, 96, 104649. doi: https://doi.org/10.1016/j.scs.2023.104649.
•  Ratti,  C. and  Biderman,  A. (2017). "From Parking Lot to Paradise". Scientific American, 317(1), 54-59. doi: https://doi.org/10.1038/scientificamerican0717-54.
•  Şen,  Z. and  Altunkaynak,  A. (2006). "A Comparative Fuzzy Logic Approach to Runoff Coefficient and Runoff Estimation". Hydrological Processes: An International Journal, 20(9), 1993-2009. doi: http://dx.doi.org/10.1002/hyp.5992.
•  Şen,  Z. and  Altunkaynak,  A. (2009). "Fuzzy System Modelling of Drinking Water Consumption Prediction". Expert Systems with Applications, 36(9), 11745-11752. doi: https://doi.org/10.1016/j.eswa.2009.04.028.
•  Sunitiyoso,  Y.,  Wicaksono,  A., et al. (2023). "Future of Mobility in Jakarta Metropolitan Area: A Multi-Stakeholder Scenario Planning". Transportation Research Interdisciplinary Perspectives, 19, 100810. doi: https://doi.org/10.1016/j.trip.2023.100810.
•  Thakur,  P.,  Kinghorn,  R., et al. (2016). "Urban Form and Function in the Autonomous Era". Proceedings of Australasian Transport Research Forum (ATRF), 38th.
• TÜİK. (2024). Retrieved from https://www.tuik.gov.tr/.
•  Van Arem,  B. (2015). "Road Vehicle Automation: So Much More Than Technology". Proceedings of TRB Annual Meeting, Washington, DC.
•  Wang,  Y.,  He,  J., et al. (2024). "Driving into the Future: Multiview Visual Forecasting and Planning with World Model for Autonomous Driving". Proceedings of Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 14749-14759.
•  Zadeh , L. A. (1965). "Fuzzy Sets". Information and control, 8(3), 338-353.
•  Zhang,  W.,  Guhathakurta,  S., et al. (2015). "The Performance and Benefits of a Shared Autonomous Vehicles Based Dynamic Ridesharing System: An Agent-Based Simulation Approach". Proceedings of Transportation Research Board 94th Annual Meeting, 2919.