Game Theory Applications in Micro and Macroscopic Simulations in Transportation Networks: A Comprehensive Review

Transportation networks are essential to the functioning of modern societies. They provide access to goods, services, and opportunities, connecting people and places. However, the efficient and effective management of these networks is a complex challenge that requires careful consideration of numerous factors, including capacity, demand, congestion, and user behavior. Game theory (GT) has emerged as a powerful tool for analyzing transportation systems and developing strategies to optimize their performance. This manuscript provides a comprehensive review of the application of GT in transportation networks from both microscopic and macroscopic perspectives. Specifically, it examines how GT has been used to analyze the behavior of individual travelers and transportation providers and how it has been used to develop strategies for managing congestion, improving efficiency, and reducing emissions. Lastly, the manuscript provides a roadmap for future research and highlights the challenges in the application of GT in transportation networks.


I. INTRODUCTION
Transportation networks are critical for the smooth functioning of modern economies. With the growing demand for efficient transportation, the focus has shifted to developing models that optimize the network's performance [1].
GT has gained significant attention in the field of transportation network research due to its potential to model the interactions between various agents in a transportation system. It is a mathematical tool that can be used to analyze strategic interactions between individuals or groups and can provide a valuable framework for understanding the behavior of agents in a transportation network. It aims to identify the optimal strategy for each player, considering the choices of the other players. In the context of transportation networks, GT can be applied to analyze the behavior of passengers, operators, and policymakers. Passengers aim to minimize their travel time and cost, while operators aim to maximize The associate editor coordinating the review of this manuscript and approving it for publication was Yanli Xu . their profit. Policymakers aim to maximize the overall efficiency of the transportation network while balancing the interests of all stakeholders [2]. GT has been applied to a wide range of transportation network problems such as traffic congestion, route choice, and parking management. These problems are often complex, involving multiple agents with different objectives, and can be difficult to model using traditional mathematical methods. GT provides a framework for modeling such complex interactions and can help to identify optimal solutions that balance the interests of different agents in the system [3], [4]. Figure 1 illustrates the general research process of GT in transportation network.
The application of GT in transportation network research can be broadly categorized into two perspectives: microscopic and macroscopic.
The microscopic perspective of GT in transportation network research considers the interactions between individual travelers and their decision-making processes. This perspective is useful for understanding the behavior of individual travelers and how their decisions can affect the performance of the transportation network. The main focus of microscopic GT is to model the interactions between individual travelers and their decision-making processes and to identify strategies that can improve the overall performance of the transportation network [5].
One of the main areas of application of GT in transportation network research from a microscopic perspective is in the modeling of traffic congestion. Traffic congestion is a significant problem in urban areas, and it can have a significant impact on the efficiency of the transportation network. GT provides a valuable framework for modeling the interactions between individual drivers and their decision-making processes, and for identifying strategies that can alleviate congestion [6]. Another area of application of GT in transportation network research from a microscopic perspective is in the modeling of route choice. Route choice is a fundamental decision that individual travelers make when navigating a transportation network. GT provides a framework for modeling the interactions between individual travelers and the different routes available to them. By modeling the interactions between individual travelers and the different routes available to them, GT can help to identify strategies that can improve the overall performance of the transportation network [7].
On the other hand, the macroscopic perspective GT can be used to analyze the overall performance of the transportation network. One area where macroscopic GT has been applied extensively is in the study of traffic congestion. Congestion is a major problem in many urban areas, leading to increased travel times, reduced reliability, and negative environmental impacts [8]. GT can be used to model the behavior of drivers in congested networks and develop strategies for reducing congestion. For example, researchers have used GT to design toll pricing schemes that encourage drivers to travel at different times or take alternative routes, reducing overall congestion [9]. Another area where macroscopic GT has been applied is in the design of transportation networks [10]. Traditional approaches to transportation planning often focus on optimizing the network based on a single criterion, such as travel time or cost. However, GT can be used to consider multiple criteria and model the interactions among different actors in the system. For example, researchers have used GT to design transportation networks that balance the needs of different stakeholders, such as residents, businesses, and commuters [11].
Over the last few decades, various researcher has tried to implement the GT to improve the possible outcomes of the decision-making behaviors in the transportation network. Despite the significance of GT in improving transportation networks, only a handful of researchers have conducted a state-of-the-art review on GT applied to transportation networks. In this context, [12] puts forward arguments regarding the potential role of GT in transportation modeling. While the authors provide an overview of non-cooperative traveler behavior, [13] delves into the application of GT in transportation analysis, but restricts its focus to micro and macro policy applications. On the other hand, [14] provides a comprehensive review of the urban traffic management system, but confines its scope to the vehicular network. Additionally, [15] surveys the use of GT in lane-changing applications, whereas [16] summarizes the existing literature on GT-based transport market modeling.
Compared to the literature, this manuscript has the following contribution.
1. This manuscript provides a comprehensive understanding of the role of GT in transportation networks, including its theoretical background, and applications. 2. The manuscript examines how GT has been applied to transportation networks at a microscopic level, including the modeling of individual agents' behavior, such as drivers or passengers, and their interactions with each other. 3. The manuscript also explores the application of GT in transportation networks at a macroscopic level, including the modeling of system-level behavior and interactions, such as the allocation of resources or the coordination of multiple modes of transportation. 4. The manuscript evaluates the effectiveness of GT as a tool for analyzing and optimizing transportation networks. This could include a discussion of the strengths and limitations of GT, as well as case studies or empirical evidence of its use in real-world transportation networks. 5. Finally, the manuscript could outline future directions and challenges in the application of GT in transportation networks, including potential areas for further research, emerging technologies, and potential policy implications.
The remainder of the manuscript is organized as follows; Section II explains the GT classifications and their applications in transportation networks. Section III explains the possible application of GT in the transportation network. Sections IV and V provide an in-depth review of the microscopic and macroscopic perspectives of GT in the transportation network. Section VI suggests the challenges and the future work while Section VII concludes the manuscript.

II. GT CONCEPT AND CLASSIFICATION
GT constitutes the mathematical examination of competition and conflict, illustrating how structural components between players can result in overall outcomes based on the choices made by individuals. Such outcomes may not have been actively sought by any of the involved parties. Games can be considered as mathematical objects, comprising a group of players, a collection of strategies or moves available to them, and payoffs associated with every potential result of the game for any combination of strategies. Players refer to the individuals or groups who are involved in the game. Each player has a set of strategies that they can use to achieve their objectives. Strategies refer to the set of actions that a player can take in the game. Each strategy is associated with a set of possible outcomes and payoffs. Payoffs refer to the rewards or benefits that a player receives for each possible outcome in the game. Payoffs are often measured in terms of utility or a similar metric [17]. The players' decisions and the type of game they play are influenced by these payoffs. If the payoffs are constant or equal to zero, the individuals have opposing interests and are engaged in a zero-sum or constant-sum game, in which one player's gain is equivalent to the other's loss. Non-zero-sum games, where the total payout is not constant or equal to zero, are more complex and in certain situations allow for greater collaboration [18].
Also, based on personal interests, GT can be utilized to predict how individuals might behave during arguments. In a standard game, decision-makers or players strive to outwit one another by anticipating their opponents' moves. The decisions made by the players determine the game's outcome [19].
Cooperative, non-cooperative, and evolutionary game theory (EGT) are the three main branches of GT, and their further classification is illustrated in Figure 2.

A. NON-COOPERATIVE GAME THEORY
In non-cooperative GT, the focus is on analyzing games in which players act independently and pursue their own selfinterest. The players are assumed to have no means of communicating with each other and must make decisions based on their own information and preferences. Non-cooperative GT deals with issues such as how to model strategic interactions, how to identify equilibria in a game, and how to predict outcomes based on players' actions. Further, the non-cooperative GT can be classified into multiple game models [20]. Ordinary Non-Cooperative Game: It is no longer tricky to create a standard non-cooperative game model. The core concept is to analyze the issue, isolate the three computing elements, construct a model, explore the attributes of the equilibrium position, and explain the difficulty [20].
Generalized Nash Equilibrium Game: Each player's choice has an impact on not just the payoffs but also the range of possible options available to them [20].
Cournot Game: It is the first application of Nash equilibrium, as well as a well-known example in GT and a specific instance of the prisoner's dilemma paradigm [22].
Stackelberg Game: In a Stackelberg game, the leader has an advantage in terms of information and strategic positioning, as they can anticipate the follower's response and adjust their decision accordingly, while the follower's decision is constrained by the leader's choice. This game is often used to model situations in which one player has a dominant position in a market or industry [23].
Game of Bounded Rationality: While GT has had a lot of success, some individuals have questioned whether the participants are rational. Because it is founded on the concept that each player wants the optimal outcome but can only earn restricted payoffs, a bounded rational game is considerably closer to reality [24].
Repeated Game: A repeated game is made up of many stages, each of which constitutes a full game. Although a recurring game is only a version of fundamental games, the end outcome might be rather different [25].
Many of the troubles associated with transportation networks fall within the boundaries of a noncooperative Stackelberg or Nash game. This research compares the two games and provides instances of issues that may be solved using either framework. It discusses the contrasts between different sorts of challenges and broadens the spectrum of potential solution strategies by emphasizing their relationship with GT [26].

B. COOPERATIVE GAME THEORY
In cooperative GT, the focus is on analyzing games in which players can form coalitions and work together to achieve a VOLUME 11, 2023 93637 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply. common goal. The players are assumed to communicate and cooperate with each other to maximize their joint payoff. Cooperative GT deals with issues such as how to divide the payoff among the members of a coalition, how to form stable coalitions, and how to allocate resources efficiently among the members of the coalition [27].
Cooperative games can be classified into two main categories: transferable utility games and non-transferable utility games. In transferable utility games, the utility or value associated with a given outcome can be transferred between players, meaning that the total amount of utility available to the players can be increased or decreased by the formation of coalitions. Examples of transferable utility games include the classic examples of the prisoner's dilemma and the chicken game. In non-transferable utility games, the utility or value associated with a given outcome is non-transferable, meaning that it cannot be split up and assigned to individual players. Instead, the value of the outcome is determined by the identities of the players who are involved in the coalition. Examples of non-transferable utility games include the games of matching pennies and the volunteer's dilemma [28].

C. EVOLUTIONARY GAME THEORY
EGT is a branch of GT that takes into account the role of evolution and natural selection in shaping the strategies and behaviors of individuals in a population. It is based on the idea that social behaviors and strategies can evolve over time, just as physical traits do in biological evolution. In EGT, players are assumed to be able to adapt their strategies over time based on the outcomes of their interactions with other players. This can lead to the emergence of certain strategies that are more successful in the long run. One of the key concepts in EGT is the notion of fitness, which refers to the reproductive success of an individual or a strategy. Fitness can be measured in terms of the number of offspring an individual has or in terms of other factors that contribute to reproductive success, such as access to resources or social status [29].
In transportation systems, EGT has been applied to analyze the behavior of drivers, the emergence of traffic congestion, and the adoption of new transportation technologies. It has also been used to study the behavior of users in wireless networks and the design of efficient communication protocols [4].
One of the main findings of EGT is that the outcome of a game depends on the structure of the population and the type of interactions that take place between individuals. In an unstructured population, where individuals interact randomly with each other, EGT tends to promote defection over cooperation. However, in organized groups, where interactions are confined by either social or geographical ties, cooperation is preferred [30].
The EGT system model consists of three components: population, game, and replicator dynamics, as shown in Figure 3. Replicator equations are a tool for studying evolutionary dynamics, depicting the rate of increase in the percentage of organisms using a certain tactic. Equations for continuous replicators assume infinite populations, unbroken time, full mixing, and the validity of strategies. The equations' attractors (stable fixed points) represent evolutionarily stable states. A strategy is considered evolutionarily stable if it can outlast all ''mutant'' strategies, indicating that it is significantly influenced by heredity and animal behavior [32], [33]. Further, the comparative difference between various forms of GT organized in Figure 2 is well presented in Table 1.

III. GT APPLICATIONS IN TRANSPORTATION NETWORK
The application of GT in road transport networks can be seen from various perspectives. One of the most important applications of GT in road transportation is traffic flow management. By modeling the behavior of individual drivers and their interactions with each other, GT can help identify optimal traffic flow strategies that reduce congestion and improve travel times. For example, GT can be used to analyze the effects of road pricing policies, where drivers are charged for using certain roads or entering certain areas during peak hours. This can help reduce congestion by encouraging drivers to shift their travel times or choose alternate routes [13].
GT can also be used to study the impact of road infrastructure investments. By modeling the decisions of different entities involved in the development of road infrastructure, such as government agencies, private investors, and construction firms, GT can help identify optimal investment strategies that balance the costs and benefits of different road projects. This can help ensure that limited resources are allocated to projects that have the greatest impact on transportation network performance [34].
Another important application of GT in road transportation is route planning and navigation. By modeling the behavior of individual drivers and their interactions with each other, GT can help identify optimal routes that minimize travel times, avoid congestion, and reduce fuel consumption. This can help drivers save time and money, while also reducing the environmental impact of transportation [35], [36].
Further, GT can be used to optimize vehicle routing and scheduling in logistics operations. By modeling the decisions of logistics companies and their drivers, GT can help minimize the number of vehicles needed to serve a given set of customers, reduce travel times, and improve on-time delivery performance. This can help logistics companies save money while also improving customer satisfaction [37].
This manuscript categorizes the GT applications in transportation systems into two categories: macro-policy analysis and micro-behavior modeling. The earlier is focused on circumstances in which a great number of participants take part in a game in a multifaceted and vast area. Either one, on the other hand, focuses on isolated circumstances in which only a few players exist in a narrow region [38]. Table 2 explains the various differences between micro and macroscopic-based GT applications in transportation networks. Further, the detailed classification of micro and macroscopic applications in transportation networks is illustrated in Figure 4.

IV. GT APPLICATIONS IN MACROSCOPIC SIMULATION ANALYSIS OF TRANSPORTATION NETWORK
The majority of scientific studies on the effects of social policy use a micro, ''history,'' and measured approach. As a result, the projections' veracity is restricted in scope. As a substitute, a macro, theoretical, and organizational perspective is offered. Such a shift in public health policy would provide not just a better understanding of the scope of policy forecasts, but also chances to test them both historically and in comparison. It would also bring together policy analysis and sociology theories [8].
Based on the fact that transportation players can be divided into two categories: government (public, civic, local associations or companionship in-service community transportation examines or toll roads) and passenger (users of transportation systems), we make the following distinction in this section: games played by travelers and officials, games played by authorities, and games played by passengers [44]. Macroeconomics is concerned with big financial phenomena that have an impact on the whole of society. As a result, policymakers must make systemic decisions like yield curve fixing and combining a state's prices with its exports and financial currency markets. Supply-side economic theories, which are basically the polar opposites of Marxist theories, say that tax increases discourage private capital and, as a result, stifle the development necessary for a growing economy. Lowering taxes, on the other hand, indicates the administration has less cash to blow, which might lead to a rise in deficits as a result of increased public debt. Policymakers are continually trying to prevent a depression, which occurs after a major recession has occurred [45].

A. ROAD PARKING TOLL ANALYSIS
Managing the flow of vehicles and crowds in transportation has become a challenge in recent times. In some cases, determining optimal prices involves not only transportation issues but also the provision of value-added services. In ITS, every service or infrastructure used, such as roads, parking lots, or public transportation, comes with a price. The need for surge pricing, demand pricing, or time-based pricing in ITS is elevated in situations where special services or value-added services are provided or when demand is high, and supply is low [46].
Road parking tolls are a form of congestion pricing, where drivers are charged a fee for parking their vehicles in certain areas. GT can be used to analyze the behavior of individual drivers and parking lot operators and to develop strategies for optimizing parking toll policies [47].
In GT, the players in a system are typically modeled as rational decision-makers who try to maximize their own interests. In the case of parking tolls, the players might include individual drivers who decide whether to park in a certain area, parking lot operators who decide how much to charge for parking, and government regulators who set policies and regulations related to parking tolls [48].
One approach to analyzing parking tolls using GT is to model the behavior of individual drivers and parking lot operators as a non-cooperative game. In this game, each player tries to maximize their own interests without considering the VOLUME 11, 2023 TABLE 1. Comparative difference between various forms of GT [4], [12], [13], [26], [36], [39], [40], [41].
impact of their decisions on the other players. For example, a driver might try to find the cheapest parking spot, while a parking lot operator might try to charge the highest price possible. However, this type of non-cooperative behavior can lead to inefficient outcomes. For example, if too many drivers park in a certain area, congestion and delays can occur, reducing the overall efficiency of the transportation network. To address this issue, GT can be used to model the behavior of drivers and parking lot operators as a cooperative game, where the players work together to achieve a common goal, such as minimizing congestion or maximizing revenue [49], [50].
Next strategy for optimizing parking toll policies using GT is to use dynamic pricing. To establish dynamic prices, it is necessary to have information on current parameters, historical parameters, and previous prices. Dynamic pricing is calculated based on different factors that can be influenced by time, demand, weather conditions, and culture. Therefore, periodically analyzing and prioritizing all the factors/parameters correctly is critical to ensuring the effectiveness of dynamic pricing models. Dynamic pricing involves adjusting parking tolls based on the level of demand for parking in a certain area. For example, if a certain parking lot is nearly full, the price could be increased to encourage drivers to park elsewhere, reducing congestion and improving the efficiency of the transportation network [47], [51].
Another strategy for optimizing parking toll policies using GT is to use market mechanisms, such as auctions, to allocate parking spaces to drivers. In an auction system, drivers bid for the right to park in a certain area, and the parking lot operator awards parking spaces to the highest bidders. This can help ensure that parking spaces are allocated efficiently and that parking tolls reflect the true value of parking in a certain area [52].
The most economically efficient fee structure is determined to be the optimum day for road tolls based on theoretical analysis and quantitative samples. When toll levels are suitably established, charging a period cost has no impact on the system of moment road tolls. The use of a duration-based parking cost regimen is found to be less effective than the system of autonomous morning commute travel plans that are free of tolls and parking charges. The common societal response decreases as the length-dependent parking fee rates increase within the framework of period-dependent parking charges. Varying driveways will result in different walking hours, which will impact commuters' travel time to the workplace, as well as their arrival percentages and time and service costs. However, in this research, commuters will not need to commute due to the fully autonomous traffic situations. Instead, they will be driven to the central city (for work), and automated cars (AVs) will return on their own (selfdriving) and pick a parking spot along the route. Previous research results suggest that applications of this kind can be further classified into two groups: games among passengers and establishment (usual techniques for traffic signal management at intersections) and games among travelers (which often occur in highly incomplete positions) [53].
As noted in [54], road pricing presents various implementation issues, including high service costs, conflicting traffic, and political resistance. Due to its non-intrusive payment method that does not disrupt traffic flow, parking charges are often favored over road tolls. Consequently, road pricing has been implemented in only a few locations, despite parking charges being collected in almost all metropolitan cities. This section focuses on analyzing a location-based parking tax regime that excludes road tolls for commuting from home to work. The analysis includes both on-and off-street parking conditions in a micro parking model. A parking methodology based on the concept of proximity is presented in the microscopic framework to simulate movement, and a trans-portation state monitoring approach is suggested to collect an event-based set of data for simulation.
The impact of parking costs on travel patterns is limited to commuting trends. Therefore, optimizing travel patterns based on nighttime travel in regime is feasible. To assess the influence of service charges on vehicle ownership, it is important to observe market pricing for on-street parking or alternatives in close proximity. Secure parking prices in various countries can be monitored due to a large property market for privately held parking. However, in many nations, privately owned parking is typically bundled with residential properties, and residents pay for parking through the purchase of the property or controlled parking permits. This makes it difficult to determine the fees for privately offstreet parking. Additionally, in areas where parking is in high demand, the fees charged for parking may include the time spent searching for a spot [55]. Parking issues are thoroughly and systematically examined to identify their causes. Finally, recommendations and alternatives are presented. Disordered parking not only affects the daily transportation of city inhabitants and their quality of life but also has a negative impact on the city's reputation.
Various researchers have applied the applied the GT, one approach is through Stackelberg models, as demonstrated in a study [56]. This model designates the government as the leader with the goal of maximizing welfare by setting the price for on-street parking, which takes into account the cost of searching for a limited number of spaces. On the other hand, the commercial, off-street operator is considered the follower and sets prices to maximize profits. By solving the model for Central London, researchers found that setting the price for on-street parking equal to that of off-street parking is a straightforward policy measure that maximizes welfare. Moreover, a straightforward approach of time restriction, where the driver is not charged if the parking is for a limited period, is less beneficial than an optimal meter fee because it leads to increased search costs. Thus, pricing is a better tool for regulation in such cases compared to quantity levers [57]. The authors of [58] investigate a system for dynamic pricing and parking reservation in parking lots. They create a Stackelberg game, where the parking agency determines the prices and the drivers decide where to park, which affects the demand distribution. They use a bilevel, multiperiod mathematical program with equilibrium constraints (MPEC), solved through dynamic programming, to generate a set of prices based on the level of occupancy. In [59], a Wardrop model is utilized to evaluate the effects of parking policies on welfare and distribution in Zurich, Switzerland. The researchers discovered that the current parking policy in the city is not effective, and while new policies could improve efficiency, they may also be regressive. However, there are two aspects that have not yet been analyzed using game-theoretic tools in real-world applications: (i) the pricing of residential parking [60], and (ii) the pricing of workplace parking, which could impact employees' mode choice [61]. VOLUME 11, 2023 [49], [50], [52], [54].
Further, Table 3 compares the Cooperative GT strategy (with toll), Non-Cooperative GT strategy (no toll), and EGT strategy (adaptation over time) for road parking. The Cooperative GT strategy aims to maximize social welfare by introducing a toll, which encourages drivers to share the cost of parking and reduce congestion, leading to reduced environmental impact and improved travel times. The Non-Cooperative GT strategy focuses on individual benefits, where drivers park for free, but this results in congestion, environmental impact, and delayed travel times, reducing overall social welfare. The EGT strategy involves the emergence of strategies that maximize individual benefits over time based on the outcomes of interactions. This may lead to the potential overuse of resources and reduced overall welfare [49], [50].
Overall, the Cooperative GT strategy appears to be the most beneficial for both the government and drivers, as it leads to reduced congestion, improved environmental impact, and decreased travel times, thereby maximizing social welfare. The Non-Cooperative GT strategy may lead to short-term benefits for individual drivers but can result in reduced overall welfare. The EGT strategy can result in the emergence of strategies that maximize individual benefits but may not be optimal for overall social welfare in the long run.

B. VEHICLE ROUTING PROBLEMS
GT can be applied to address various challenges in vehicle routing problems (VRPs), such as coordination among vehicles, competition among customers, and uncertainties in demand and traffic conditions [62].
One application of GT in VRPs is the cooperative game approach, where vehicles are treated as players in a cooperative game. The objective is to minimize the total travel time or distance for all the vehicles collectively. In this approach, vehicles coordinate with each other to find an optimal routing solution that benefits all of them. The challenge is to design a fair and efficient cost-sharing mechanism that incentivizes all vehicles to cooperate [63].
Another application of GT in VRPs is the non-cooperative game approach, where customers are treated as players in a non-cooperative game. The objective is to maximize their own utility, such as minimizing the waiting time or travel cost. In this approach, customers compete with each other for the services of the vehicles. The challenge is to design a mechanism that incentivizes the customers to reveal their true demand and preferences, and that results in an efficient allocation of vehicles [64].
GT can also be used to address uncertainties in demand and traffic conditions in vehicle routing problems. For example, stochastic GT can be applied to model the uncertainty in demand and traffic conditions as random variables and to design routing policies that are robust against such uncertainties. This approach involves finding a strategy that maximizes the expected utility under all possible realizations of the uncertain variables [65].
To address the static route planning issue, the approach divides it into several static vehicle routing problems, with the ultimate goal of optimizing customer loyalty. The approach is centered around turning on-the-road cars into virtual consumer points. The first instance of a dynamic vehicle routing issue is discussed in [66]. With the logistics industry evolving rapidly in the direction of global expansion and digitalization of operations, logistics distribution has become a critical link between suppliers and users, especially across the distribution chain. To this end, the authors examine a single-vehicle DARP in which client queries are dynamically produced as they drive from one location to another. Their method takes advantage of insertion heuristics, which perform effectively with minimal processing work.
Subsequently, [67] introduced the concept of instantaneous request, wherein a customer in need of help always desires to be supplied as quickly as possible, requiring an urgent re-routing of the present vehicle route. The main task of optimizing the distribution model is to optimize distributed vehicle planning, which includes freight collecting lines, product transmission systems, and the convergence of commodities gathering container loading, and dispatch.
Extensive research has been conducted on the VRP and other related dispatch issues, as they are both practically and theoretically interesting. The design of routes for trucks conducting distribution or activities is of substantial business interest since it directly correlates with cost-cutting. The fact that this generalizes the Travelling Clerk Problem while being vastly more challenging has kept theoreticians interested in it for years. Varieties of the VRP occur due to practical issues such as vehicle capacity, delivery time frames, road system delays, and the ability to divide shipments [68].
Several technological developments have increased the range of actual routing applications. With the advent and widespread use of Global Positioning System, mobile and smart phones, and dependable Geographic Information Systems, companies are now able to efficiently monitor and manage their vehicles in real-time at a minimal cost [69]. The argument for reducing range was perhaps that fuel usage is proportional to the distance traveled. However, various additional variables influence fuel use, and several writers have subsequently concentrated on decreasing overall fuel consumption while recognizing load carried and/or speed. Acceleration's major impact on gasoline usage, however, has not been taken into account.
In the past, vehicle routing was carried out in a two-stage process of planning and execution. However, modern technological advancements have made it possible to perform routing dynamically, which offers numerous benefits such as reducing operational costs, improving the customer experience, and minimizing environmental impact [70]. VRPs are more complex than static routing problems and offer additional degrees of freedom that pose new challenges in evaluating the value of a particular routing plan [71].
On-street parking increases aren't necessarily a good thing, as a high transport vacancy rate reduces vehicles' chances of obtaining a parking spot, lengthening cruising durations, and degrading internet traffic performance. Low parking utilization, on the other hand, is wasteful in terms of space usage. With differential driving, the situation becomes even more problematic. This example shows how dynamic routing changes routes on the fly, requiring actual communication among trucks and the dispatching station, where the surroundings refer to the real world and the dispatcher is the agent responsible for delivering orders to the vehicle [9]. Further, Table 4 provides comparative analysis of the application of GT in VRPs In VRPs, the choice of the GT approach will depend on the specific circumstances and goals of the problem at hand. For example, if the goal is to minimize overall transportation costs and all players are willing to cooperate and share resources, cooperative GT may be the most appropriate approach. However, if there is competition between players or if communication or coordination is not possible, non-cooperative GT may be a better fit. EGT may be useful in situations where strategies can evolve over time based on the outcomes of interactions, such as when optimizing routing algorithms or improving traffic flow. Ultimately, the choice of approach will depend on the specific problem and the assumptions that can be made about player behavior and interactions.

C. TRANSPORTATION NETWORK RELIABILITY AND TRAFFIC FLOW ANALYSIS
GT has been widely applied to the transportation network reliability problem to help identify and mitigate the effects  [62], [63], [65], [72], [73].
of system disruptions, such as accidents or weather events, on the overall performance of the network. In this context, GT models the interactions between different agents, such as commuters, transit operators, and infrastructure managers, to evaluate how their decisions and actions affect network reliability [74].
One key application of GT in transportation network reliability is the study of the ''user equilibrium'' problem, which involves modeling the behavior of individual commuters in selecting their routes in a congested network. By using GT to model this behavior, we can identify the optimal set of routes that minimize travel time and improve network reliability. Another application of GT in transportation network reliability is the study of collaborative decision-making processes among different agents in the network. By using GT to model these interactions, we can identify strategies and policies that encourage cooperation and coordination among different agents to improve network reliability [75] Furthermore, GT has also been used to analyze the impact of disruptions on transportation networks, such as accidents, weather events, or infrastructure failures. By modeling the interactions between different agents in the network during a disruption, we can evaluate the effectiveness of different response strategies, such as rerouting traffic, deploying emer- VOLUME 11, 2023 93643 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply. gency services, or implementing temporary infrastructure changes [76].
The transport system can operate under normal or deteriorated conditions, which may arise due to incidents [77]. The probability of the network's reserve capacity meeting or exceeding the required consumer capacity, despite the degradation, is referred to as the probability of reserve capacity [78]. Combining capacity and travel time reliability can be a useful tool for network design. By incorporating network reliability principles in the planning and evaluation of traffic management solutions, researchers can guide the development of effective driver information management and traffic control systems [79].
Relative accessibility, which represents the level of connectivity between two places or a specific location and a movement, can be measured in terms of travel distance, journey time, or transit cost. This concept is commonly used to determine the location of emergency services such as fire, ambulance, or police, where the proximity of the nearest facility is critical [80]. The relative accessibility B jk between two places j and k may be expressed as Eq. (1).
where, D jk represents the gap between the places in proportion to the above-mentioned three terms. Regardless of network vulnerability, this description of accessible examines the defeat of the usefulness of transportation among two nodes as a result of network deprivation. A connective vulnerability is a name given to this kind of vulnerability. Integrated availability, on either side, displays the connections between a certain location and all other places or behavior inside an area. In general, integral accessibility (BJ m ) is determined by adding the relative accessibility of all locations k to the relevant availability of place J, as explained in Eq. (2).
As a result, integrated availability may be evaluated using a simple gauge of travel partition, as given by equations (1) and (2), which is acceptable for applications such as seeking urgent situation help. However, the statistics do not take into account the level of service supply at any given site, nor does it provide for user selection of a specific facility. Composite metrics of integrated availability that include global supply parameters and geographic distance have been devised, with the Hansen index [81] being two such measures. Further, explained in Eq. (3).
Second, the Black-Conroy cumulative distribution index, indicated by, Third, the Black-Conroy cumulative distribution index, represented by, where g(D jk ) is an impedance function and B k represents the quantity or strength of activity at k, given in Eq. (4).
where BJ j (U ) represents the total number of chances across U travel cost units of location J and O j (u) represents the objective number of chances across U travel cost units of location J [81]. Although it has the benefit of providing an incessant evaluation of the available surface from a single location, this index has been sparingly used, most likely due to the random choice of the restricting trip cost units CD. This surface may be used to evaluate the relative availability of a certain place across an area, and dissimilar J ideals would result in surface contours. Variations of these contours in various network degradation scenarios might serve as the basis for monitoring overall information about the network susceptibility and the comparative belongings for different positions within the network.

D. URBAN TRAFFIC DEMAND
Theoretically, public transit primarily operates on a small portion of urban roads. The fast traffic speeds of these services highlight the need to divide lanes and optimize the number of junctions and roads for improved network usage. Compared to other factors, the impact of lane width on service traffic volumes is considered significant. The findings can be useful for urban planning authorities and responsible organizations [82] and many method for urban traffic demand prediction including forecasting one can be developed. The generic process to predict the urban demand is well illustrated in Figure 5.
Forecasting models for travel are utilized to anticipate alterations in travel behavior and the use of transportation systems that occur as a result of modifications in regional development, demographics, and transportation supply. Estimating travel demand is a difficult task, but it is essential for logical planning and assessment of transportation systems. Decision-making in transportation planning involves identifying potential upgrades to a community's roadway infrastructure. To help in this process, various computer-based and manual tools have been developed [83].
The application of a specific game model is proposed to determine urban traffic demand and address underlying difficulties, emphasizing the need for the government to consider the desires of traffic players and management efficiency, along with imposing differentiated fines on rebellious drivers. Developing nations allocate less than 10% of land for road building, whereas the corresponding proportion for rich countries ranges from 20% to 35%.
To reduce transportation's environmental impact, cities must promptly adopt the Russian Association's transport policy (Russian Federation, 2011). The main objective of this research is to establish a model of sustainable urban development or ecological citizen movement. Vuchic divides transportation planning into four levels, with the first level outlining the link between the municipality and transportation, 93644 VOLUME 11, 2023 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply. specifically the interplay between transport networks and other city elements. The study's main aim is to design a paradigm for sustainable urban development or environmental citizen engagement [84]. In the study by the authors in [85], two matching game models are compared for addressing traffic congestion issues, and a Nash equilibrium is identified. The authors further demonstrate that in the absence of road pricing, roads would be excessively utilized due to the principles of human rationality, which provides a theoretical foundation for traffic congestion pricing. Collaborative ramp combining has emerged as a popular solution to this problem, facilitated by the progress in connected and automated vehicle (CAV) technology. CAVs can not only communicate with each other in a coordinated traffic setting, but also manage complicated situations involving humandriven vehicles. Reference [18] employs a differential game model to elucidate the dynamic traffic obligations and traffic signal control, establishing its effectiveness in a basic traffic network. As urbanization continues to accelerate, the transfer of freight to, from, and within metropolitan areas has emerged as a critical concern. While industrial transportation is crucial for urban retailing and manufacturing, it has significant adverse impacts on urban quality of life in terms of traffic flow, pollution, and space utilization. Urban mobility programs have advocated for collaborative and environmentally conscious urban transportation as a means of alleviating the negative consequences of urban mobility, but these initiatives face organizational and technical challenges. The paper [1], gathers and analyzes current advancements to collaborative transporting goods in urban traffic environments. Transferring freight from, to, and inside metropolitan areas has become a key concern as the world becomes more urbanized. Further, Table 5 presents the comparative analysis of the application of GT in predicting urban traffic demand

E. TRANSPORT MODES COMPETITION
The world economy has entered a new phase with the global economic slowdown and sluggish global trade. The construction of terminals has become increasingly challenging. However, there has been steady growth, particularly in small and medium vehicles, filling the gap between marketplaces held by major urban areas. Despite extensive studies on the concept, construction, and management of transport networks, little research has been conducted on the geographical planning of large transport infrastructure while taking into account competitors in the market [88]. VOLUME 11, 2023 93645 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply.  [48], [86], [87].
Recent studies suggest that road competition occurs at both the strategic and functional levels. The decisions made at each level interact with each other, and the simultaneous consideration of both levels can maximize the profits of players. Each mode of transportation has unique operational and economic characteristics and benefits. However, modern demand requires interconnected transportation networks, which necessitate flexibility in the use of each mode. Consequently, modal competition arises in various degrees and across several dimensions, including price, efficiency, availability, regularity, security, convenience, and more. To complement each other, specific modes need to meet three major requirements [89].
Different Geographic Mark Place: In the context of multiple markets, it becomes necessary to ensure continuity within the transportation system, especially when dealing with different scales, such as national and international transportation. This requires the establishment of an interconnection, commonly known as a gateway, to enable the transition from one mode of transportation to another. Intermodal transportation has played a crucial role in enhancing complementarity and connectivity across various geographical markets [16].
With the widespread use of connected devices, travelers rely on socially regulated taxi services based on factors such as geographical location and advertised fares in social taxi communities. Traditional taxis or privately owned vehicles may also be utilized. However, the popularity of this automotive components calling system has presented a number of challenges, including the problem of territorial distribution. Social cab drivers tend to concentrate in areas with the highest concentration of potential customers, resulting in a negative impact on each other's profits due to high competition [83].
To address this issue, [90] proposes a collaborative territorial distribution strategy that can be implemented through an agreement among telecom operators. The territorial distribution problem is formulated using behavioral economics and can be solved using a concessions project specific. The solution model is intended to be a no-regrets game with a coarse coupled stability as the outcome.
Various Transportation Markets: In transportation systems, the mode of transit utilized for either people or goods often indicates a certain degree of complementarity. Although the same market may be supplied, it may not be equally accessible based on the mode of transportation employed. Hence, in certain markets, rail, and road transit may complement each other, with one focusing on people and the other on freight [16]. Despite extensive research on supply chain management, the sustainability of oil and gas distribution chains has not received adequate attention. ''Sustainable development is the structured collaboration of critical cross-functional and cross-business operations for enhancing the long-term economic productivity of individual operations and supply chain, while achieving an association's strategic, ecological, and stated objectives,'' defined Carter and Rogers. This notion drives governments to establish necessary regulations on transport options and facilities to reduce environmental pollution and simultaneously increase profit. The concept of a Responsible Petroleum Supply Chain aims to generate employment and income for each participant while concurrently reducing CO 2 and other carbon emissions. Despite its importance, this concept has received limited attention [91].
Various Degrees of Service: Two modes of transportation that offer distinct levels of service are likely to complement each other by providing specialized offerings for a comparable market and accessibility. The most common complementarity is cost versus time. Inland transportation and multimodal transportation systems have significantly improved in accuracy due to consolidation and information systems in recent years. Consequently, each harbor has a significantly greater reach into its hinterland, and the markets that various ports can access, and service are increasingly overlapping, causing port rivalry to intensify. The response of rival ports will have an impact on the investment of a port and its regional economy [92].
Modal competition occurs when there is an overlap in geography, transportation markets, and degree of service. The most important factor in mode selection is cost, which is determined by the price/performance profile of each mode and the distance traveled, quantity transported, and value. In order to address the transportation challenges of historic areas, various game concepts have been developed with 93646 VOLUME 11, 2023 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply.
adequate data, taking into account financial, ecological, and traffic growth considerations. These concepts are designed to maximize overall profitability while modeling the impact of environmental degradation, energy consumption, and road service levels on traffic flow. By repeating the Nash equilibrium result of the system, the optimal structure and proportion of transport modes in historic areas can be anticipated. Therefore, historical and cultural preservation must always take priority in the design and construction of these areas. In contrast, urban areas are often caught in a destructive spiral of traffic congestion due to ineffective road traffic management. However, this approach is costly and can seriously damage the historical cityscape and ecosystems, putting the natural system and transit systems on the verge of collapse. As a result, finding a way to reconcile insufficient portability with the ever-increasing demands of overall traffic and structural diversity to create a new quality of system connection has become an urgent issue [16], [88], [89].
In order to promote the use of sustainable transportation methods such as walking, cycling, public transit, and carpooling, a more effective and long-term approach is required to address urban traffic issues. This approach can also contribute to the development of a peaceful and sustainable urban environment. Research on urban traffic planning has identified four key areas for investigation. The first area involves understanding the relationship between urban spatial configuration and traffic patterns. Studies have shown that urban land use patterns and density have a significant impact on travel behavior, and researchers have identified the factors that influence vehicle travel speed and distance using descriptive and analytical methods. In addition, experts have proposed various urban traffic planning ideas that consider land use patterns, traffic structure, population density, and city size based on the characteristics and development stage of urban areas [93].
In recent years, there has been a growing trend toward using optimization techniques to design transportation systems that consider multiple goals. The third area of research focuses on developing optimization models for urban traffic structures, which can be approached in one of three ways based on prior research findings. Various urban traffic planning recommendations have been proposed from a land use perspective, taking into account the transportation system, urban form, population size, and urban size in accordance with urban land use characteristics and traffic patterns [94].
On other hand, Transport mode choice is a crucial element of transportation planning, which is a challenging task due to the complexity of the decision-making process involved. Several game models have been developed to analyze transport mode choices and predict travelers' preferences. In this comparative analysis, we will review and compare different game models used in transport mode choice.
Nash Equilibrium Model for Transport Mode: The Nash Equilibrium Model is one of the most widely used game models in transport mode choice. It assumes that individuals choose their preferred mode of transportation based on their own benefit and without considering the actions of others. The model predicts the outcome of individual choices, where each individual tries to maximize their utility, given the choices of others. The model assumes that the market is in equilibrium when no individual has the incentive to change their mode choice [95].
Stackelberg Model for Transport Mode: The Stackelberg Model is another popular game model used in transport mode choice. In this model, one player, known as the leader, makes a decision before the other players, known as the followers, and the followers' decisions are based on the leader's decision. The model is useful when one player has more power or influence than others. The leader determines the mode of transport to use, and the followers choose the mode that maximizes their utility, given the leader's decision [96].
Evolutionary Game Model for Transport Mode: The Evolutionary Game Model is a relatively new approach to transport mode choice. It assumes that individuals learn and adapt their behavior over time based on their experience. In this model, individuals start with a randomly selected mode, and they switch to another mode based on their own experience or feedback from others. The model predicts the long-term distribution of mode choices based on the individual's experience and the interactions between individuals [97].
Bayesian Game Model for Transport Mode: The Bayesian Game Model is a more advanced game model used in transport mode choice. It assumes that individuals have imperfect information about their preferences, the choices of others, and the outcomes of their choices. The model predicts the probability of individuals choosing a particular mode of transport based on the information available to them. The model is useful in situations where individuals may have incomplete or uncertain information [98].
Further, Table 6 presents a detailed comparative analysis between game theoretic and non-game theoretic methods for transport mode choice.

F. RISK ALLOCATION
The risk allocation ratio is dependent on a series of interchanging offers, the cost of capital, and the asymmetric intensity of information in cases where the negotiation process was initiated in the first round with either the public or private sector. It is evident that the risk allocation ratio is closely associated with the sequential manner of interchanging offers, the cost of capital, and the degree of information asymmetry [99].
GT has been increasingly applied to transportation networks to allocate risks between public and private sector stakeholders. Transportation infrastructure agreements involve the process of risk distribution between the public and private sectors, which can be viewed as a bargaining process between these two players. A final offer arbitration game is used to represent this procedure [100].
Transportation systems are critical to the functioning of vital systems since they allow quick access to other assets and VOLUME 11, 2023 93647 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply.  [83], [95], [96], [97], [98].
relief efforts during catastrophes and throughout the recovery phase. As a result, improving the resilience of transportation systems and understanding the complexities of rehabilitation behavior is critical, and these can be assessed and analyzed using potential options for a quick recovery [101].
The application of GT in risk allocation aims to evaluate the behavior of the players when presented with opposing goals in risk allocation. The model demonstrates that when the value of assurances exceeds the value of a financial loss, strategic behavior and possible moral hazard issues may arise. Quantitative risk distribution approaches are being developed to address the constraints of qualitative risk distribution methods, particularly the question of the amount of risk to be shared by each party. Table 7 illustrates the various game models used for risk allocation.
In transportation infrastructure projects, most risks can be legitimately allocated, but demand risk remains vague owing to the many variables determining its proper allocation. Therefore, using indicators as tools to assist contractual parties in allocating demand risk is essential. The use of GT in risk allocation can assist in addressing concerns such as the amount whereby the parties share risk difficulties. Special methods are restricted, and quantitative risk distribution approaches are being developed to address these constraints [102].
According to infrastructure agreements related to transportation [104], the distribution of risk between the public and private sectors is regarded as a negotiation process. To represent this process, a final offer arbitration game is utilized. Since transportation systems facilitate quick access to assets and relief efforts during and after a disaster, they are essential components of critical systems. Therefore, improving their resilience and comprehending the complexities of recovery behavior is vital, and possible recovery options can  [99], [100], [102], [103].
be assessed and analyzed. The aim of this study is to examine player behavior in risk allocation with conflicting goals using a game framework. The model indicates that when the value of assurances is greater than the value of the financial loss, strategic behavior and potential moral hazard issues arise [105]. Traditional methods are limited in addressing concerns such as the distribution of risk between the parties. Thus, quantitative risk distribution approaches are being developed to overcome the limitations of qualitative risk distribution methods, specifically addressing the question of how much risk should be shared by each party [106].

V. GT APPLICATIONS IN MICROSCOPIC SIMULATION ANALYSIS OF TRANSPORTATION NETWORK
Micro behavior simulation refers to the computerized modeling of individual vehicles, drivers, and their interactions within a transport network. This approach involves simulating the movement of individual vehicles and their interactions with other vehicles and the infrastructure, such as traffic signals, road signs, and road geometry. Micro behavior simulation is commonly used to evaluate the performance of a transport network under different conditions, such as changes in traffic volumes, road designs, or traffic management strategies [107]. By modeling the behavior of individual drivers and vehicles, this approach can provide a detailed understanding of the impacts of changes on traffic flow, travel times, and congestion levels. Based on earlier research findings, these applications can be categorized into two groups: games between travelers and officials, which generally require techniques for traffic signal regulatory oversight at intersections, and games among travelers, which regularly 93648 VOLUME 11, 2023 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply. take place in a very partial space [108]. This section provides detailed game theoretic development in various microscopic simulation-based applications as defined in Figure 3.

A. ADJACENT TRAFFIC SIGNAL STRATEGY
The strategy of adjacent traffic signals is linked to agent collaboration, coordination, and information exchange [109], [110], which can be used to alleviate traffic congestion. Since traffic signal operators are spread across the road network, teamwork among them can lead to optimal scheduling. GT can represent a connection with several intersections, with each unit representing a signal regulator that can adapt to changes in the external environment. The complexity arises from the fact that the actions of one operator affect the surroundings and are influenced by the behavior of other operators [111]. Although single agents (players) are self-interested and seek to maximize their gains, this leads to suboptimal junction strategies. Therefore, for real-world applications such as traffic signal control (TSC), information exchange is necessary to make the best decision. Various methods have been developed to deal with TSC, and a comparative literature of a few of them is explained in Table 8.
The authors in [112] present a methodology for reducing traffic congestion at an isolated junction using GT and a Markov chain model. The increase in traffic and lack of available capacity in the highway system poses threats and issues for shippers, passenger transportation, water quality, and human safety. Traffic congestion leads to increased fuel consumption, emissions, and pollutants, affecting global warming and air pollution. The reduction of traffic congestion enhances passenger accessibility and lowers fuel consumption and emissions. However, actuating and adaptable regulators are limited by min and max cycle duration, green indicator timings, delays, and pre-defined patterns of stages. Algorithms that centralize decision-making increase vulnerability to collapse, making scaling up difficult, comparatively more complicated to run, and more expensive. In a noncooperative game, each intersection is treated as a player at a different crossroads, with each contributor aiming to reduce its queue length. Nash equilibrium is used to determine the optimal signal-timing technique for a signal controller, and Nash bargaining is used in [113] to evaluate the optimum TSC method at a specific intersection.
Previous research focused on a single intersection, but it is essential to consider the effectiveness of a traffic network with multiple intersections. Urban road congestion leads to air and noise pollution, and loss of time for travelers, and has significant societal impacts. Several traffic management techniques have been developed to address this problem, such as increasing the number of roads and widening them. However, these methods do not effectively improve urban traffic adaptability [114]. The use of stochastic GT and reinforcement learning was investigated to optimize the coordination between two intersections [115]. Traffic signals affect people's daily lives in significant ways, impacting drivers, the environment, and the economy. To achieve network-wide performance objectives, traffic engineers, manufacturers, and academics have proposed the use of Traffic Signal Timing (TST) systems, which coordinate individual traffic lights. The concept of agents has been applied to TST, and features such as distribution, autonomy, and collaboration are well-suited to the congestion domain. Scholars have proposed a variety of methodologies, including GT, machine learning, fuzzy logic, and supervised learning, for scheduling signalized intersections. Intersection agents are depicted as players who make choices based on both their local interests and a global objective.
The authors of reference [116] employed a Markov-chainbased GT approach to model the synchronization of multiple traffic signal controllers through simulation. TSC models for highways, trains, and other transportation systems have evolved from simple pre-timed solitary indicators to more complex automated and linked TSC models. However, previous TSC systems may still encounter problems such as congestion caused by weather conditions, delays caused by accidents, and overloading. Hence, proposing a TSC solution model for multiple junctions that can modify traffic signal timing based on actual traffic is a significant challenge.
Ensuring efficient travel from origin to destination is one of the most important challenges in mobility science, and traffic signal management plays a critical role in addressing this issue. Improving signal timing management for a region to reduce overall wait time is a challenging subject. The objective is to reduce the time spent waiting at a particular junction. If each signal controller selects its colors based on its own local traffic considerations, road traffic will be disrupted because the traffic flow will not take into account the other signal controllers. However, even in a non-cooperative strategy, if every indicator connects with others simultaneously and modulates its stages in response to changes in global traffic flow, congestion in different parts of the city can be highly optimized. Intelligent traffic systems have attracted the interest of both academicians and practitioners as a means of reducing traffic jams. Artificial intelligence approaches, particularly those involving reinforcement learning, are suitable for identifying such methods, as they use various inputs and do not require explicit analysis comprehension or modeling of the basic dynamic systems. The authors proposed a method for efficiently calculating optimal system behavior. The method consists of two steps: estimating the equilibrium position and fine-tuning. They used nonlinear programming to develop the game.
In order to manage traffic lights, a combination of fuzzy Q-learning and GT was utilized in [126] to simulate the effects of agents' actions on their neighbors. With urbanization and an increase in vehicle flow coupled with a lack of mass transit, journey time, fuel consumption, and air quality have all been impacted. Traditional transportation planning methods are ill-equipped to handle the complexity and ever-changing nature of large traffic systems. Intelligent traffic systems are increasingly incorporating artificial VOLUME 11, 2023 93649 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply. intelligence techniques. In these systems, cars report their average waiting time to the traffic light, which uses a common algorithm known as Q-learning to control traffic signal programming.
Reference [127] investigated the application of GT in a two-level TSC system. The lower level is concerned with user equilibrium, while the upper level considers green-splits of signals as decision-making factors. The study proposes to leverage Internet of Things infrastructures to optimize traffic flow in transportation systems. In urban settings, intersections and linkages constitute the bulk of the traffic infrastructure, leading to frequent traffic congestion and associated problems. Eliminating these obstacles not only improves travel experiences but also addresses several issues such as environmental pollution.
Urban road networks typically consist of a mixed system of two types of intersections: signalized and nonsignalized. Signalized intersections use traffic lights, while non-signalized intersections do not. In reality, both types of intersections are present in urban road networks, leading to various traffic volume patterns and characteristics. To improve traffic flow, traffic signal controllers are expected to be intelligent agents capable of collaborating with each other to form a coalition game. EGT is useful in proactive decision-making. A novel agent architecture is created using EGT to improve TSC and signal controller cooperation. The Internet of Things has impacted various industries, including transportation. The concept of ''Smart Transit,'' based on IoT technology, is considered a future direction for the transportation sector. However, the rapid proliferation of connected devices has resulted in an increasingly complex and fragile ecosystem, making negotiation strategies for agent cooperation a popular topic in the field. Among the vari-ous techniques, Smart Traffic Light Control (STLC) is the most promising for improving traffic flow management as it can be applied to different traffic settings. The STLC challenge involves multiple intersections cooperating and sharing data to reduce road congestion and improve vehicle wait times [128].

B. LANE-CHANGING
The ongoing futuristic transformation of the transportation era indicates that the coexistence of manually driven and autonomous vehicles is expected soon [129]. If autonomous vehicles behave differently from human drivers, traditional drivers might become anxious and distrustful. Consequently, road safety would be further compromised. Improper driving practices cause the majority of traffic accidents. A safe and robust road network is primarily influenced by how drivers operate their vehicles. Imitating a driver changing lanes (to the right and left) and turning (to the right and left) frequently causes the most accidents in a traffic flow and is known to have an adverse effect on traffic flow stability [129], [130], [131]. Figure 6 illustrates the lane changing (LC) scenario under aggressive and cautious competing vehicle [132]. With advanced driver assistance systems, including blind spots and lane departure warnings, one may be able to reduce the risk of accidents caused by lateral movements to a certain extent. However, in the real world, the ability of the driver to use turn signals intuitively is highly significant. Reports state that turn signals are used at a rate of 40% in China, while they are used at only 44% in the United States [133], [134]. Therefore, these systems do not significantly impact the lives of most drivers. At the same time, they do not solve the problem of frequent accidents in which lateral movements are a contributing factor. 93650 VOLUME 11, 2023 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply.

1) METHODOLOGICAL DEVELOPMENT IN LC
The two most common methods drivers interact with traffic flow are changing lanes and following other vehicles. Unlike car-following, LC behavior tends to be more complex. For the first time, the authors in [135] developed the lane-changing decision model for learning how to navigate. This model was developed to understand the factors that may influence drivers' decisions to change lanes in urban traffic and establish a logical framework for making lane-changing decisions. As part of an extension of Gipps' original model of lane switching on the expressway and urban roads, researchers have examined the likelihood-based lane changing [136] aggressive driving behavior [6] and adverse effects such as accidental collision and traffic congestion [7]. Some scholars, however, have attempted to improve Gipps's model by using various structural equation models. According to [137], they developed the ''Minimizing Overall Braking Induced by Lane Changes'' model, which considered changes in acceleration of the vehicle in addition to the original model. In an effort to implement lane-changing strategies on freeways utilizing cooperative merging algorithms [138] and optimization algorithms [139].
Furthermore, another stream of scholars examined LC behavior using utility theory. It was the first time utility theory was applied to assess the LC process in a study on freeway lane change decisions [140]. Reference [141] included MLC and DLC behaviors in their utility model. Using driver profiles and vehicle trajectory data to analyze the effects of different driver physical characteristics on LC behavior, [142] and [143] have discussed how driver characteristics influence the LC behaviors of the other drivers.
Lane changers have the role of decision-makers, serving as the central unit of analysis within the LC decision and utility theory model. A problem with this approach is that no consideration is given to the reactions of passing vehicles. The lane changer and the driver in the adjacent lanes can be affected by one another during lane changes. Lane changers' behavior is partly influenced by the actions of the previous vehicle and the action of the vehicle moving in the adjacent lane. An LC maneuver also affects the fellow's driving behavior and the leading vehicle's driver [144].
A research effort needs to be carried out on building algorithms capable of recognizing human decisions and interactions to solve these problems. Based on these algorithms, models will be developed that can be used to predict drivers' actions. Based on the analysis of interactions between players using GT or mathematical models, such algorithms can be  created. Based on the various GT-based LC models developed in the last two decades, is illustrated in Figure 7.
From various perspectives, researchers have used GT to study the problem of lane-changing on highways, a generic GT based LC approach is shown in Figure. 8. In the methodological development of the GT based LC models, known as a pioneer, Kita [145] was the first to use a GT approach to analyze lane-changing behaviors. As a result, the researchers' community investigated several factors, including traffic conflicts with drivers in interweaving areas [146], enhanced merging process [147], traffic safety [148], [149], [150], and driving space requirements [151] in addition to comfort requirements [152]. Additionally, new driving environments, such as the connected vehicle environment [133], [148], [149], as well as the automated driving environment [134], are gradually being considered among LC players. As far as game models are concerned, researchers have applied pure strategy games [153], repeated games [154], mixed strategy games [146], Stackelberg games [155], [156], cooperative games [157], [158], [159], EGT [160] and Harsanyi theory [148], [149].

2) RECENT DEVELOPMENT IN GT-BASED LC MODELS
Initiative approach of the [145], with perfect information, merging and giveaways based non-zero-sum, noncooperative games between two players, made it possible for GT-based LC models to be developed in the future. Further, in line [161] developed a pay-off estimation method to improve the equilibrium selection problem. This model analyzes merging giveaway behavior with the new concept of ''probability of equilibrium selection''. In this study, Kita has investigated the optimal selection of equilibrium in multiple equilibrium scenarios and developed a method to determine each player's payoff function and the optimal equilibrium selection probability. Thus, the equilibrium selection criteria no longer need to be defined a priori. A field study using observation data about merge-giveaway behavior at on-ramp merging sections revealed that the proposed model demonstrated reliable prediction capabilities of equilibrium selection probability and that drivers' interactions were reasonably well described with the model. Kita's model [161] has certain limitations, and Liu [147] developed an enhanced GT-based model for modeling the merging of vehicles. The speed of vehicles is modeled as a series of variables rather than a set constant in Kita's model to capture realistic behavior. Behavioral rules are also incorporated into the payoff functions within this classic Nash equilibrium model. The authors also implemented an estimate method that utilizes a dual-level valuation method; the upper level solves a least-square problem while the lower level deals with a linear complementarity problem. An LC behavior model in a connected environment [148] has employed two different games to predict LC behavior. In the first case, the authors have considered an incomplete information-based ''two-person nonzero-sum non-cooperative game'' according to Harsanyi's principles [162], where complete information is transformed into imperfect information with both DLC and MLC taking into consideration, in the second, a complete informationbased ''two-player non-zero-sum non-cooperative'' games in which vehicles communicate vehicle-to-vehicle in order to receive more accurate information in the future, enabling them to perform lane-changing maneuvers that are more reliable and safer.
The authors have used the revised version model of stochastic Nishinari-Fukui-Schadschneider in [163]; the model considers the influence of the distance between vehicles and various driving behaviors, like starting slowly, accelerating quickly, or braking randomly. Further, the authors formulated four actions based on two players' games using both the prisoner's and social dilemma. Reference [151] has enhanced the previous LC model (discussed in [163]) to a multi-player (three players) based Stackelberg game. According to this study, GT-based systems are more likely to be effective in enabling better lane changes and outperforming systems driven by fixed rules. The entire process of changing lanes resembles human behavior: first, the host vehicle uses the turn signals or makes a quick lateral move to interact with the surrounding vehicle; the competing vehicle responds accordingly, e.g., accelerates or decelerates; finally, by determining the rival vehicle's aggressiveness and the pay-off function, the optimal timing and acceleration are determined. The GT-based MLC model introduced in [149] has been successfully applied in a traditional environment. The lane-changing model has not been extensively explored in a connected environment, so they extended it to a connected environment. In order to evaluate the performance of the system, factors such as location prediction and time prediction error, false alarm rate, detection rate, and true/false are considered. Likewise, [164] modeled the simulation as an incomplete information-based ''two-player non-zero-sum'' non-cooperative game. However, in modeled game, the subject vehicle has only two options; waiting or merging while the following vehicle has four possible ways; changing the lane, not changing the lane, decelerating, or accelerating.
A model was developed afterward in which failed DLC attempts were used to reduce traffic interruptions and ensure traffic safety [133]. In the lane-merging scenario, the authors in [165] have introduced a Stackelberg game model that examined the payoffs associated with vehicle aggressiveness. Such models could be helpful in traffic control and autonomous driving. In lane-change maneuvers, authors of [129] have applied a Stackelberg GT model to determine the steering torque. The electric power system and drivers are taken as two players in the modeled game. Based on the Stackelberg game principle, the proposed model has developed a steering energy-saving pattern considering electric power system as a follower and the vehicle driver as the leader. The obtained results show that compared to Nash's strategy, Stackelberg's strategy has the potential to save drivers more maneuver forces.
Authors in [166] have proposed a Nash Q-learning-based motion decision algorithm with surrounding vehicles interaction prediction capability which determines the exact timing and actions to execute the LC process. Further, the proposed work has introduced a physical method-based game during the surrounding vehicle's trajectory prediction. The result has shown improved comfort and safety. However, this work failed to incorporate the action recognition of surrounding vehicles, reducing candidate trajectory numbers and improving the decision process efficiency.
Using the concept ''internet of vehicles,'' the authors in [167] have introduced a novel algorithm called ''multivehicle coordinated lane change'' for MLC decisions. In the proposed algorithm, the vehicle motion is optimized using 93652 VOLUME 11, 2023 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply.
the model predictive control method, and a Stackelberg game-based relation is modeled between the surrounding and the subject vehicle. The obtained results under different scenarios show that the developed algorithm improves traffic conditions of the roundabout by increasing the average speed while reducing the vehicle acceleration fluctuation. However, the model has ignored the communication delay, as well as high-speed vehicle moment, hence more efficient algorithms and control methods need to develop.
The authors in [51] have developed a cooperative LC strategy based on GT. The authors have formed a single-vehicle LC strategy by introducing the comfort advantages into velocity advantages, while the multi-vehicle cooperative LC strategy is explored using a double-game matrix. Simulation results have shown improved driving comfort during the LC process. However, considering uncertainties due to external interference that affects the vehicle decision analysis needed further research.
A novel LC model based on a multi-player, non-zero-sum, non-cooperative sub-game has been developed by the authors in [168], where each vehicle is considered a player. Each cell within the traffic map corresponds to a sub-game, where the vehicles act based on solutions to Nash equilibrium problems generated by the sub-game. This method aims to plan a route schedule that leads vehicles to approach the target lanes as many times as possible, depending on the time available. The obtained results are compared with the gap acceptance strategy. The comparative analysis shows that the proposed LC model is more efficient.
Modifying the conventional method, a hybrid (multi types, i.e., car, bus, truck, etc.) condition vehicle-following model is developed in [169], covering both LC and vehiclefollowing conditions. Further, a highway-based two-vehicle game model of DLC decision is created. An optimal strategy is determined by considering a comprehensive loss function, including efficiency, comfort, and safety. [170] has presented a fast and reliable GT-based optimal decision-making framework for the LC process. The method is quick as Pair-wise games are effortlessly solved by ego. Then these solutions are integrated into a final decision which reduces the required computational time to solve a multi-player-based N-dimensional game. Using an intelligent algorithm, the parameters for the game were tuned and players' behavior was adjusted accordingly. A simulation shows that the proposed decision-making strategy is advantageous since it considers the objectives of other agents before selecting an action. However, the proposed methodology is only capable of solving Ego vehicle decision-making. Further, efforts must be made to solve global optimization problems that account for driving scenarios like intersection crossings and other decision-making variables such as fuel economy.
Considering the passenger's comfort, power performance, and driving safety, the authors in [55] have proposed a Bézier curve path planning-based GT approach for the LC decision-making process. LC data of 83 drivers are used to obtain the LC safety distance. The safety consideration gets enhanced after considering the LC time and safety distance calculated by the path planning layer into the game payoff. The closed-loop feedback from the planning layer to the decision layer improves the security of decision results. Next in the proposed model, a constrained optimization methodology is developed, which further enhances the planned path's comfort, traceability, and safety. Results obtained during the Hardware in Loop testing show the superiority of the proposed methodology in terms of LC time and vehicle interval control and provide an optimal balance between traceability and safety. However, the proposed method has not considered multiple driving scenarios, indicating further research on efficient algorithms with the capability to solve more complex problems.
Further, the pay-off function and equilibrium selection were also estimated using simulated moments [171], even though there was no explicit definition of the pay-off function. It is pertinent to note that these are only two-players games, while the actions have been extended up to three; accelerating/decelerating and changing lanes, etc. By computing the time-to-collision calculation, the final maneuver of merging and through cars, equilibrium can be determined. However, multiple equilibrium states, constant speeds, and interactions only among neighboring vehicles (e.g., over a small-scale area) could deliver impractical results. Considering the drawbacks associated with the classical GT such as issues with equilibrium selection, hyperrationality, and lack of dynamicity. The concept of the EGT for the LC application came into the picture. Relatively few studies have been carried out on the EGT-based LC model. The EGT applied to LC models provides a general explanation of the cooperative interactions between drivers from the perspective of an entire society. However, the researchers concluded that integrated tools might negatively affect cooperative collaboration. The use of simulators to develop cooperation strategies has been the subject of several recent studies. Further research should be conducted regarding the feasibility of weakening or eliminating dilemma effects.

C. EMERGENCY RESCUE
In the case of transportation system emergencies, such as accidents, natural disasters, or terrorist attacks, multiple agents are involved in the rescue operation, including emergency responders, transportation operators, law enforcement, and civilians. Each of these agents has different goals, incentives, and constraints that can lead to conflicting decisions and outcomes [172].
GT can be used to model the interactions between these agents and predict their decision-making behavior under different scenarios. For example, GT can be used to analyze the strategic interactions between emergency responders and transportation operators in deciding which routes to take or which resources to allocate to the rescue operation. It can also be used to analyze the incentives of civilians in fol- lowing evacuation orders or cooperating with the rescue operation [80].
In the last few decades, researchers have developed various models to analyze interactional behavior as well as outcomes during emergency rescue. Table 9 illustrates a few of them, used for emergency rescue in transportation networks.
The authors of [172] proposed a framework to minimize travel time for urgent transportation network design, using the branch-cutting approach and indicators such as travel duration, intensity, and routes to identify optimal paths for urgent vehicles. The study aims to develop a comprehensive framework for locating emergency facilities in transport systems, considering various models such as place and position models. The primary emergency transportation network is identified through the emergency transport network design problem and managed by the police for essential journeys and disaster aid activities. Other sections of the system may also be used for emergency aid, but designated primary areas are regulated due to their importance. Common journeys may use the main network but avoid channels designated for disaster response travels.
In [175], the authors performed path-searching to determine expected travel time in dynamically changing traffic conditions and employed a flexible signal assumption to mitigate its negative impact on the overall traffic flow. They also introduced elastic signal preemption, which involves solving quadratic programming challenges to devise non-intrusive signal timings for emergency vehicles to pass intersections without stopping. In real-world scenarios, the ESP module reduced response time by approximately 30%, while not affecting the overall traffic situation.
In reference [176], urgent situation evacuation paths are evaluated in emergency situations, where factors such as distance and traffic are taken into account. However, the impact of roadway traffic control information on route selection is not considered. The penalty cost of a method limitation is measured in terms of satisfaction. To account for uncertainty, the Bertsimas robust optimization approach is utilized, and the decision-making strategy for the receiver path is adapted to mitigate volume fluctuations [177] Proposed a reliable route-finding approach for a diverse road network with unpredictable travel times that is applicable in real-world traffic networks. Battery electric vehicles (BEVs) have become increasingly popular among passengers and organizations due to their environmentally friendly features. However, the underlying infrastructure supporting BEVs is still in its early stages of development, leading to operational difficulties and inefficiencies for BEV users.
In [178], the issue of incorrect traffic state predictions and improper guided route determination induced by the deceptive traffic state data generated by cooperative vehicles is addressed, which results in a secure traveler information process and enables traffic efficiency and safety. Some tourists or drivers may not consider the time cost when selecting a route but instead focus on the shortest path length from the start point to the destination. The Transportation-based Cyber-Physical System or Intelligent Transportation System employs advanced communication devices to enhance the efficiency, reliability, safety, and resilience of transportation systems. Real-time route planning systems that combine actual traffic status with route maps to help passengers identify optimal paths for their journeys have been developed to save time for travelers and alleviate congestion problems.
Furthermore, various authors (refer to Table 10) have formulated various game models with the intent to maximize the payoffs/utility function under emergency evacuation.

D. CONFLICT BETWEEN TWO VEHICLES
The emergence of automobiles, especially at intersections, is expected to transform future transportation management. Several studies have proposed a distributed conflict-free cooperation mechanism for many linked cars at unsignalized junctions. Despite the growing need for mobility, road transportation faces significant obstacles. The popularity of the internet of autonomous cars has risen due to its potential to enhance transportation safety [187]. CAVs form a multi-agent network with three-layer and interconnected characteristics. Compared to normal urban roadways, junctions pose a more complex and challenging problem for multi-vehicle synchronization. The double vehicle formation at the junction, where automobiles enter through various entries, cross over their position trajectory at the junction zone, and depart through different outputs, adds to the complexity. The intricate conflict connection between cars in this double configuration makes it difficult to prevent crashes at the junction. Moreover, vehicle construction at a junction with a large number of 93654 VOLUME 11, 2023 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply.  different cars exhibits high-dimensional and discontinuous dynamics. Therefore, achieving multi-vehicle cooperation at junctions is challenging [188]. It begins by projecting incoming cars from various traffic motions into a virtual lane and then incorporates a conflict-free geometric topology that takes into account the conflicted relationship of engaged vehicles, resulting in the formation of a virtual platoon. Figure 9 illustrates the representation of the vehicle's traffic assessment.
In reference [189], the authors have utilized model predictive control to supervise the potential safety conflict risk and fuse decision-making with overtaking maneuver control. This approach's effectiveness has been established in various experimental settings, and the effects of specific factors on this control method have also been examined. The study presents a unique approach for directing multiple vehicles from given start and destination configurations while avoiding collisions. The policy assumes that all agents cooperate by adhering to the same traffic rules. Although centralized traffic control problems may have precise and complete solutions, they frequently necessitate significant processing power.
Furthermore, centralized methods are often susceptible to decision-maker errors. Decentralized methods require each vehicle to plan its own path based only on locally available information, such as the location and direction of nearby vehicles. A decentralized approach is often quicker in responding to unforeseen problems, but safety certification is a concern since the cascading effects of potential conflicts may impede converging to solutions in certain circumstances. Additionally, a decentralized architecture ensures the system's sustainability. The approach suggested in [190] is applicable to systems where new vehicles may enter the scene at any time and begin interacting with existing ones, while others may depart.

E. COLLISION AVOIDANCE BETWEEN VEHICLES
GT can be applied in collision avoidance between vehicles to develop effective strategies for safe driving. GT involves modeling the interactions between drivers as a game, where each driver is a player and the strategies, they use to avoid collisions are their moves. The goal of each player is to avoid collisions while reaching their destination in the shortest possible time [191].
In this context, GT can help to identify the Nash equilibrium, which is the point where no player can improve their strategy by changing it unilaterally. By analyzing the Nash equilibrium, drivers can identify the best strategies to avoid collisions and reach their destination safely. Various GT techniques can be used in collision avoidance between vehicles, such as the Prisoner's Dilemma game, the Chicken game, and the Stag Hunt game. These games can help to identify the optimal strategies for avoiding collisions and reducing the risk of accidents. Table 11 illustrates the various game models from the perspective of collision avoidance between vehicles.
Some other research has implemented deep reinforcement learning with and without integrating the GT. Deep Reinforcement Learning (DRL) is an approach that integrates a learning algorithm with deep learning. This technique enables an agent to acquire knowledge on how to operate within a given environment through the use of feedback incentives and penalties. The related functional form is approximated using a neural network, allowing the suggested architecture to assess various actions depending on an innate comprehension of the events. This enables the agent to learn from scratch without VOLUME 11, 2023 93655 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply. the need for human guidance or expertise, thereby potentially facilitating the discovery of exceptional behaviors and enhancing performance. The remarkable accomplishments of spanning gaming algorithms are driving more academics to explore practical applications of DRL. For instance, a DRL method that employs off-policy training of deep Q-functions effectively acquired a challenging door-opening skill on an actual robot manipulator [195].
Untimely collision avoidance systems are focused on managing obstructions by detecting potential hazards using decision trees and utilizing route planning to avoid obstacles. Computer vision plays a critical role in local routing and is a fundamental aspect of both vehicle and occupant safety. Over the years, several studies have been conducted on this topic, and numerous solutions have been proposed, but only a few have been applied in practical systems [19]. To avoid collisions, a robot must not only identify impediments but also calculate an optimal trajectory and navigate safely in real-time. Unmanned aerial vehicles (UAVs) equipped with sensing and communication systems have recently gained significant attention due to their widespread availability and applications. One of the biggest challenges in automated vehicles is guidance, which comprises two parts: global path planning and localized collision avoidance. Local collision avoidance employs a specified pointer allocation to avoid obstacles, while global navigation involves developing a series of checkpoints from the starting point to the destination, taking into account the barriers in the workplace [196]. Since the 1970s, local path planning has received significant attention, and for large UAV groups, these methods are a good starting point. Multi-UAV trajectory planning techniques have been successfully employed, and multi-UAV groups are becoming increasingly common. Approaches to multi-UAV group dynamics and path-tracking management have been developed to deal with complex situations, and dispersed frameworks for multi-UAV platforms are gaining momentum. The importance of collision avoidance is driven by safety regulations in air traffic management, shipping transportation, and multi-unmanned aerial vehicle systems, where accidents must be avoided at all times. Due to their ability to ascend and descend vertically, rotary-wing UAVs are used in a wide range of applications. Various types of UAVs with different architectures are constantly being developed, and numerous techniques have been proposed to address changing barriers in real-world settings. For example, in [197], the authors present a perception collision prevention technique based on the required effort instruction; in [197], the authors explained a reactionary prevention technique based on nonlinear differential geometric guidance; and in [198], the researchers present a perception prevention technique method based on minimum attempt supervision.
Path planning is one of the most challenging tasks in multi-UAV platforms, wherein one UAV navigates a path to the destination by exchanging data with others. Due to the touch controls of UAVs, an accident-prone route may emerge, prompting the development of several obstacle detection systems. These tactics can be broadly categorized into non-cooperative and cooperative avoidance approaches. Non-cooperative strategies lack prior knowledge of communication barriers, whereas cooperative approaches rely on cooperative communication among UAVs to exchange information. For instance, passive sensors are used in non-cooperative approaches to detect and avoid obstacles, while another non-cooperative avoidance method proposed in [199] assumes that the obstacle's detachment, instructions, and rapidity variation are known. Cooperative approaches are preferred in multi-UAV systems for personal protection and design flexibility. All collision avoidance algorithms in [200] are cooperative. Most of the current techniques are implemented at the modeling layer, utilizing a combination of video sensors, long-and short-range radar sensors and other sensors integrated into the vehicle's autonomous emergency braking (AEB) system. The parameters of the sensor were derived using the specifications of a real commercial product. A flowchart illustrating the V2V communication based AEB system is shown in Figure 10.

F. CONFLICT BETWEEN PEDESTRIANS AND VEHICLES
The safety of pedestrians is a significant concern in the context of safe driving. Pedestrian areas are particularly hazardous due to the presence of crossovers. In order to examine the occurrence of pedestrian-vehicle conflicts, various methods have been employed, including the use of different conflict indicators. One such approach, described in a study by [201], involves categorizing pedestrian-vehicle traffic incidents as traffic conflicts, crucial events, and uninterrupted passages, and identifying the Post Encroachment Time (PET) as the most appropriate metric for detecting conflicts. The focus of the report is specifically on traffic conflicts involving pedestrians, which are distinguished from other types of interactions. The PET was chosen as the conflict predictor because most pedestrian-vehicle interactions involve crossing paths. In another study, [202] uti- lized the Time to Collision (TTC) as a conflict indicator to evaluate the severity of conflicts. They determined that temporal violations by pedestrians at intersections were the primary cause of the high frequency of pedestrian-vehicle incidents.
Simulation modeling of road user behavior has been recognized as a potential method for investigating road behavior in various settings. Implementing a reliable model for studying highway user activity in common space areas could prove to be a valuable tool for researchers and designers to analyze the provision and safety of common space facilities for motorists [203]. In one study, researchers investigated pedestrian conflicts with left-turning automobiles (in lefthand driving settings) in order to develop an integrated model that accurately depicted changes in left-hand-driving traffic movements at signalized crossings. The researchers used the Post-Encroachment Time (PET) and vehicle speed to verify the findings of the model. It is important for observers to properly identify conflicts between pedestrians and cars in various interactions. Historically, the threshold-based technique has been utilized to identify conflicts, as exemplified by [204], who focused solely on traffic conflicts with Time to Collision (TTC) values smaller than 3 seconds for conflict analysis.
On the other hand, [205] utilized a behavior-based technique in combination with a threshold-based strategy to identify conflict occurrences and their intensity. Traditional traffic studies relied on collision information, which often failed to provide full or precise details and neglected to indicate the underlying causes of crashes. In their study, the researchers found that different conflict indicators, such as Time to Collision (TTC) and Gap Time (GT), contributed differently to signaling the severity of conflicts under diverse interaction patterns. As depicted in Figure 11, a pedestrian may encounter other pedestrians, oncoming traffic, the crossing border, and traffic lights while passing through a crosswalk. The surrounding pedestrians include both opposing and leading pedestrians.

VI. CHALLENGES AND FUTURE TRENDS
Although, as shown in the summary, GT has determined many obstacles in transportation research, there are still disadvantages to utilizing different game replicas to explain real situations. The important issue is that the majority representations presume that all agents are balanced. Only under these circumstances can the idea be finely executed. This hypothesis, however, is not realistic since players are almost all limited and reasonable, and they are influenced by many other factors and cannot obtain all the important information. Another challenge is that it is difficult to create and solve a proper model due to the vast size of a complicated system with a large number of participants.
• Most of the previous approaches based on GT fail to analyze computational complexity since they formulate two-or three-player games with available choices of two or three. However, in real-life scenarios, the number of participating players is quite large. Thus, the computational complexity of these games analyzing the computational complexity for large systems is necessary.
• Additionally, the calculations that must be performed after constructing a game model are often rather complicated. Based on the disadvantages listed above, we may restrict our inquiry to the following areas: Use bounded-rationality game types to explain circumstances in transportation research, since this may more closely portray real-world issues.
• When the contract with big schemes, try to categorize the participants (particularly passengers) into various groups to make modeling simpler.
• The next significant challenge should be how to enhance present algorithms or build a new way to resolve game models.
Apart from the above, many questions remain unanswered and require further research, such as how to treat emergency trips, identify all market segments, address privacy issues, and consider equity implications of rejections. Additionally, it is essential to explore the responses of rejected travelers, establish fair pricing/priority mechanisms, and tailor them to rural and urban areas and different countries.

VII. CONCLUSION
As GT is increasingly being used in transportation analysis, it is critical to provide an assessment of these implementations to establish a general research approach and choose a path for future study. This review article provides a comprehensive overview of the role of GT in transportation networks. The manuscript examines the application of GT at both microscopic and macroscopic levels, including the modeling of individual agents' behavior and system-level behavior and interactions. The effectiveness of GT as a tool for analyzing and optimizing transportation networks is evaluated, including a discussion of its strengths, limitations, and real-world case studies. Additionally, future directions and challenges in the application of GT in transportation networks are outlined, including potential areas for further research, emerging technologies, and policy implications. The article also highlights ethical considerations, such as the potential impact of GT on equity and social welfare. Overall, this manuscript provides a valuable resource for researchers, practitioners, and policymakers interested in the application of GT in transportation networks.