Disaster Risk Reduction in a Manhattan-Type Road Network: A Framework for Serious Game Activities for Evacuation

Abstract

The increasing number of natural and man-made disasters registered at the global level is causing a significant amount of damage. This represents one of the main sustainability challenges at the global level. The collapse of the Twin Towers, Hurricane Katrina, and the nuclear accident at the Fukushima power plant are some of the most representative disaster events that occurred at the beginning of the third millennium. These relevant disasters need an enhanced level of preparedness to reduce the gaps between the plan and its implementation. Among these actions, training and exercises play a relevant role because they increase the capability of planners, managers, and the people involved. By focusing on the exposure risk component, the general objective of the research is to obtain quantitative evaluations of the exercise’s contribution to risk reduction through evacuation. The paper aims to analyze serious games using a set of methods and models that simulate an urban risk reduction plan. In particular, the paper proposes a transparent framework that merges transport risk analysis (TRA) and transport system models (TSMs), developing serious game activities with the support of emerging information and communication technologies (e-ICT). Transparency is possible through the explicitation of reproducible analytical formulations and linked parameters. The core framework of serious games is constituted by a set of models that reproduce the effects of players’ choices, including planned actions of decisionmakers and travel users’ choices. The framework constitutes the prototype of a digital platform in a “non-stressful” context aimed at providing more insights about the effects of planned actions. The proposed framework is characterized by transparency, a feature that allows other analysts and planners to reproduce each risk scenario, by applying TRA and relative effects simulations in territorial contexts by means of TSMs and parameters updated by e-ICT. A basic experimentation is performed by using a game, presenting the main results of a prototype test based on a reproducible exercise. The prototype experiment demonstrates the efficacy of increasing preparedness levels and reducing exposure by designing and implementing a serious game. The paper’s methodology and results are useful for policymakers, emergency managers, and the community for increasing the preparedness level.

1. Introduction

At the global level, the number of natural and man-made disasters and their relative consequences are increasing. In 2023, there were 399 disasters related to natural hazards. These caused 86,473 deaths and affected 93.1 million people [1]. The collapse of the Twin Towers, Hurricane Katrina, and the nuclear accident at the Fukushima power plant are some of the most representative extreme disaster events that have occurred at the beginning of the third millennium [2]. In its Agenda 2030, the United Nations (UN) recommends the achievement of Sustainable Development Goals (SDGs) strictly connected with disaster risk reduction (DRR). Among these, SDG 1 regards end poverty in all its forms everywhere, SDG 3 concerns good health and wellbeing, SDG 9 is connected to industry, innovation, and infrastructure, SDG 11 aims to make cities and human settlements inclusive, safe, resilient, and sustainable, and SDG 12 pursues responsible consumption and production [3]. In particular, this research is in line with the indicators “1.5.3 Number of countries that adopt and implement national disaster risk reduction strategies in line with the Sendai Framework for Disaster Risk Reduction 2015–2030” and “1.5.4 Proportion of local governments that adopt and implement local disaster risk reduction strategies in line with national disaster risk reduction strategies”. Progress on these indicators can be measured through the achievement of quantitative targets set at the international level, which commit states to specific actions [4].

Actions must be taken to reduce the effects of disasters in the risk cycle. The cycle consists of the mitigation and preparedness phases, to be defined and implemented before the instant when a disaster occurs, and the response and recovery phases, to be implemented during and after the disaster event [5]. This paper focuses on the mitigation and preparedness phases, which aim to support and test emergency plans through training exercises. Emergency plans outline the actions to be taken in an emergency to limit the damage caused by disasters. The plans also identify the individuals involved (planners, managers, operators, analysts, and end-users) who should implement the planned measures. However, simply drawing up a plan without testing it in practice will not have the desired effect. This depends on the potential gaps between the plan and its outcomes in their implementation [6]. It is, therefore, necessary to reduce the separation between the plan and its implementation through preparedness actions. It is a question of raising awareness of emergency situations so that managers and the community as a whole can adopt the right behaviour in line with the planning phase [7,8].

In the United States (US), preparedness activities have been codified and made standard. The Homeland Security Exercise and Evaluation Program (HSEEP) [9] has defined a set of technical guidelines for designing, conducting, and evaluating training and exercise actions. HSEEP addresses the preparedness activities at a national level, providing suggestions for local communities based on lessons learned, risk assessment, and corrective actions. Figure 1 presents the entire set of exercises proposed by FEMA in the United States. The set of exercises is classified into the following categories:

• Discussion-based, including activities realized in a “non-stressful” environment aimed at increasing knowledge of plans and procedures defined to face emergency conditions; this class comprises seminars, workshops, tabletops and games.
• Operation-based, including activities realized in a real environment, aimed at experimenting with planned procedures to be implemented in an emergency condition; this class includes drills and functional and full-scale exercises.

This paper focuses on the game exercise, which represents the most advanced level of the discussion-based class, in terms of complexity and capability. The main objective of the game is to understand and manage all the components of risk in a non-stressful context, and then to take action to reduce it. In the following, the exposure risk component will be focused upon. This component is related to the number of people present in an area where the effects of disasters are significant. The game requires the use of advanced methods, models and tools to simulate a potential evacuation scenario. The main references are gaming platforms, which allow the representation of reality with a high degree of accuracy. In this context, it is possible to consider serious games (SGs), which are training and educational activities carried out in a digital platform [10]. This type of game necessitates the development of technological platforms that integrate digital and modeling components. The literature indicates SGs that address different objectives of disaster risk management [11]. Digital platforms enable the representation of virtual and immersive realities and allow observation of the user’s behaviour in the gaming environment [12]. Some examples of SG for simulating evacuation processes are available in the literature.

To obtain information about the existing scientific literature, an analysis was carried out using the Scopus database, combining the following keywords: “Serious game” AND “Evacuation”. A total of 93 search results were found. Within these 93, the following categories of papers were excluded: papers that did not deal with evacuation as a main topic; editorials; papers on the non-open access problem where quantitative formulations are not referred to in the abstracts. Only 49 of the initial 93 papers passed these filters. Of these, only nine completely or partially referred to an outdoor environment. Of these nine, one was a literature review. The eight remaining contributions are summarized below (Figure 2).

The results of the selection show a prevalence of studies concerning SG evacuation processes in indoor environments. Only a limited number of papers concern evacuations in the outdoor environment (e.g., a city or a portion of a city), and this is the direction of this research.

The selected papers show a majority of studies involving cities or portions thereof. Tsunamis represent the majority of the events studied. Hawthorn et al. [13] provide a review of digital SGs aimed at testing communications in the event of tsunamis. The authors highlight the importance of communication and learning processes of risk phenomena, which can be carried out through exercises. To support these processes, it is necessary to develop SGs that represent the mobility of people in evacuation conditions. Kolen et al. [14] describe “SPOEL”, a serious game to simulate the mass evacuation process, implemented with the aim of limiting the consequences of a flood. With the development of information technology it has been possible to improve the level of representation of emergency phenomena, also with the aid of virtual reality (VR) tools [15,16,17]. Three of the nine “outdoor” papers ([15,18,19]) do not report the use of transport system models (TSM). The other papers refer to system components, i.e., demand, supply, or their interaction ([14,16,17,20]) but do not report quantitative formulations. The function is only expressed explicitly in one case [21], but the work concerns pedestrian mobility.

Table 1. Selected scientific works.

Disaster Event	Environment	User’s Platform	TSM		ICT	Risk	Digital Support	Reference	Year
			Component	Explicit Function
Earthquake/tsunami	Island	U	D	Yes	N/A	V, E	PC	Van den Berg et al. [20]	2018
Earthquake	City	U	N/A	N/A	N/A	V, E	VR	De Fino et al. [15]	2023
Nuclear	City	U	D, S	Yes (D)	N/A	E	PC	Yang et al. [21]	2020
Nuclear	Island	DM	N/A	N/A	N/A	E	PC	Schueller et al. [18]	2020
Nuclear	City	DM-U	DS	NO	N/A	O, V	VR	Bourgais et al. [16]	2024
Tsunami/Flood	City	DM	D, S	N/A	N/A	O, E	PC	Kolen et al. [14]	2011
Tsunami/Flood	City	DM-U	N/A	N/A	Digital media	V, E	PC	Fisaini et al. [19]	2022
Tsunami/Flood	City	U	D	No	N/A	O, V	VR	D’Amico et al. [17]	2023

DM: Decisionmaker; U: user; D: demand; S: supply; DS: demand/supply interaction; O: occurrence; V: vulnerability; E: exposure; N/A: Not Available.

reports a summary of the selected scientific works.

An analysis of the available open access technical literature highlights an international evolution in the field of preparedness and the role of emergency drills. One example above all is the aforementioned approach adopted by FEMA in the USA [9,22]. Issues related to organizational aspects are explored in depth, with reference to the parties involved, their respective roles, and the chain of command provided for in the plan. However, the reports published by FEMA over the years do not illustrate models for estimating risk reduction following the implementation of emergency drills.

From the literature analysis, the following two gap factors emerge: (1) the available SG are “black boxes” and, in most cases, they are developed only to increase users’ awareness; (2) it is difficult to use SGs to simulate real contexts, especially for technologically advanced urban areas where emerging information and communication technologies (e-ICT) tools for mobility are present. These factors imply the need to develop a transparent and open framework that combines TSM approaches with e-ICT tools.

Concerning planning and exercising activities to reduce the exposure risk component, it is necessary to model the user’s travel behaviour in an evacuation condition. The literature highlights the role of transport risk analysis (TRA) [23] to estimate occurrence and vulnerability risk components and transport system models (TSMs) [24] to estimate exposure, with evacuation being the main leverage. TSMs allow analysts to simulate demand [25], supply [26], and their interactions [27] in evacuation conditions; e-ICTs improve users’ choices using the most advanced devices. Therefore, e-ICTs need to be introduced in TSMs with specific advanced models ([25,26,28]). The challenge of this research is to integrate the potential of TRA, TSMs, and e-ICTs to support an SG exercise simulating an evacuation plan and thereby contribute to reducing exposure. Compared to classical games, SGs require a higher level of complexity, but at the same time have the characteristics needed to increase capability. This depends on the potential contribution of the combination of e-ICTs and TSMs adapted to represent how users may react and behave in emergency scenarios.

Serious games represent an evolution of the consolidated game concept. However, there is no unique definition of this evolution. The term “serious” is used to emphasize the purpose of transferring a message or knowledge, a skill or generic content to the participants. Some authors distinguish SGs from video games (or entertainment games) [29] based on their different purposes [30]. This is mainly related to the goals and intentions of the game developers. One of the most widely accepted definitions of SGs [31] combines different activities referred to the following sectors (Figure 3):

• “Experience”, concerning activities that have a practical relevance in reality, by including classic gaming activities from board games to sports that imply physical exertion;
• “Multimedia”, including hardware and software tools produced by the gaming industry, combining text, graphics, animation, audio, and tactile elements, including activity simulators aimed at training and usable in different subject areas; the main aim is to increase the immersiveness level to reproduce real phenomena in a virtual way [32];
• “Entertainment”, regarding applications that combine digital tools, generally used for fun, and practical activities (e.g., sport), including technologies developed in the field of computer science, computer tools have been created that reproduce sporting activities for “entertainment’ purposes”; the literature presents a large set of video games in the emergency field developed for these purposes (e.g., SimCity) [33].

By focusing on the exposure risk component, the general objective of the research is to obtain quantitative evaluations of the exercise’s contribution to risk reduction through evacuation, contributing to the overlap of the two gap factors present in the literature. In this context, this paper has the following specific objectives:

• To analyze the general characteristics of a serious game for supporting emergency exercises;
• To explore methods and models useful for analyzing an urban area in an emergency situation that can be integrated for supporting an SG built to simulate a plan for urban risk reduction;
• To identify the main elements of a transparent framework of a serious game combining TRA, TSM, and e-ICT tools.

To achieve these objectives, starting from a general classification of serious games and their applications, the paper specifies the proposed SG’s framework for simulating emergencies that imply a road transport evacuation.

According to the research’s objectives, the paper is organised into the following sections after this introduction. Section 2 describes the main models to introduce in a serious game for simulating an emergency. Section 3 specifies the functions of serious games for supporting evacuation exercises, presenting the main results of a prototypal experimentation. Section 4 discusses the results obtained, and Section 5 closes the paper with future perspectives of this research.

The proposed framework and results produce benefits for policymakers, emergency managers analysts, and single users.

Policymakers involved in the risk management process will benefit from the proposed framework in relation to the possibility of supporting ex ante evaluation of potential effects produced by planned procedures by means of simulation modeling.

Emergency managers can be supported through optimized emergency planning strategies, which also contribute to enhancing disaster prevention and mitigation awareness, as well as strengthening community capabilities.

Analysts increase their knowledge about the phenomenon of people’s mobility in extraordinary conditions. The framework allows analysts to examine user’s behavior and the influence of e-ICT in travel choices during an evacuation.

Finally, the proposed framework produces relevant benefits in increasing the preparedness level of community. With the support of a platform that implements the framework, single users, independently from their economic status, can better know how to save their own lives and the lives of those dependent on them.

2. Methods and Models

The general elements of an SG recalled in the introduction have to be specified in order to support transport risk analysis for estimating risk in an urban area. The focus is on the quantification of exposure risk components. This requires the construction of transport system models that can be updated by e-ICT (Section 2.4).

The SG for risk reduction must have an initial TRA phase to assess the occurrence and vulnerability levels of a given event (Section 2.1). Given the characteristics of the emergency scenario provided by the TRA, the TSM phase must allow for the estimation of mobility in evacuation conditions and, therefore, exposure levels (Section 2.2). This is useful during planning to test the effects of alternative scenarios (Section 2.3). Finally, in the e-ICT phase, the SGs must allow for the identification of the use of advanced devices and information for users, such as to improve user choices. The quality of model estimations can be improved by using these data (Section 2.4).

2.1. Transport Risk Analysis

The literature presents analytical formulations for quantifying disaster risk. Classical approaches estimate risk with functions that calculate the probability of the event occurring (occurrence) and the damage produced (magnitude). With specific reference to the transport system, the literature illustrates a range of approaches for transport risk analysis (TRA). In the most general case, the analytical formulation to quantify the risk (R) is as follows [34,35,36]:

R = O·M = O·V·E

(1)

where the terms are defined as follows:

O is the disaster event occurrence or the probability that a disastrous event appears;

M is the magnitude related to the event’s effects; these variables combine

V that is the vulnerability of a system in relation to the “incapacity” of each component to resist the different events’ effects;

E that is the exposure, or the quantity of people and goods present in the area affected by an event’s effects.

According to [7], serious game exercise, as with other exercises activities, contributes in a decisive way to risk reduction. “Experiential” games can be used to simulate the choices of those who plan, command, and control the emergency (such as planners, analysts, and emergency managers) and the single user who “undergoes” external events (natural phenomena, emergencies, security). This topic is linked to the nature of the interaction between the user and the game environment. The increased awareness of decisionmakers and users obtained through the simulations conducted during the SG produces a quantitative reduction in the risk and its components.

States with a P_min emergency plan comprising a series of actions to reduce the risk and the effects of which are to be tested and simulated with the SG. The objective is to move from a risk level in a condition where no emergency plan is available, P_not, denoted by R(P_not), to an optimal plan level, namely P_min, which determines the minimum level of risk denoted by R(P_min). There is a gap between the phase when planners define the actions of P_min and the phase when these actions must be implemented by users for responding to emergency situations. Exercises contribute to reducing this gap, by designing, discussing and experimenting with the planned actions. In the context of discussion-based class, a generic serious game (SG_i) constitutes an advancement with respect to the classical games by introducing the support of TRA, TSMs, and ICT.

The construction of a serious game aims to estimate the effects produced in an urban context determined by an extreme event, where the users do not know the planned action: this can be considered equal to the situation where a plan does not exist (P_not). At the same time, in each serious game exercise (SG_i) that simulates the effects of the generic plan, the analyst estimates the potential risk reduction R(SG_j). Then, it is possible to estimate the user’s capability to implement planned actions and then to reduce the distance from the condition characterized by the optimal plan (P_min). This can be expressed with the following expression, valid for the generic SG_j exercise:

R(P_min) ≤ R(SG_j) ≤ R(P_not)

(2)

This paper focuses on the exposure risk component (E) given in Equation (1); this implies the need to increase the knowledge about the mobility of people in emergency conditions, as recalled in the next sections, using SG.

The value of E depends on evacuation procedures and, thus, on the functioning of the transport system. This system is characterized by interactions between vehicles and between them with the transport infrastructure, considering that the infrastructure network can be strongly modified by extreme events.

2.2. Transport System Models

The exposure measurement defined in Equation (1) requires the quantification of the number of people present in the area affected by the effects of the emergency event, as well as the times and paths taken to reach safe places. To obtain this quantity, it is necessary to use transport system models which comprise a set of analytical formulations representing transport demand, supply and their mutual interactions. With regard to the ordinary conditions, transport demand is defined as the total number of journeys undertaken within a specified timeframe, taking into account factors, such as origins, destinations, purposes, modes of transport, and routes. Demand is influenced by the performance of the transport supply in relation to available infrastructures and services. The measurement of demand is determined by the level of service attributes. The performance of transport systems, in terms of traffic flows and travel times, depends on the demand/supply interaction [24,37].

Under extraordinary conditions defined by an emergency, the components of the transportation system assume different configurations for risk management strategies. Incorrect management of the transportation system during emergency conditions has the potential to generate undesirable effects, including a deterioration in performance (flows and times) and an increase in exposure.

The general analytical formulation of the supply models comprises the following equations:

f = Δh

(3)

g = Δ^t c

(4)

c = Γ (f, w, α)

(5)

where the terms are defined as follows:

c and g are the link and path costs vectors, including travel and waiting times;

f and h are the link and path flow vectors;

w is the vector of the functional characteristics of transport infrastructures and services;

Γ is the link-cost flow function vector;

Δ is the link-path incidence matrix;

α is the vector of the supply parameters.

The general analytical formulation of the demand models includes the following equations.

• The route utility function is as follows:

v_i = −ψ_i g_i ∀i

(6)

where the terms are defined as follows:

ψ_i is utility-scale parameters in the route choice model, for the o-d pair i;

g_i is the block of the vector of total path costs, for the o-d pair i;

• The route choice function is as follows:

p_i = p_i (v_i; β_i) ∀i

(7)

where β_i is the vector of calibrated parameters.

The system of supply and demand equations represents the supply–demand interaction.

The estimation of mobility under normal conditions is possible using models formulated in scientific literature to study transport systems [24,28].

The transport demand/supply interaction can be studied using analytical models (Equations (3)–(7)) (with Γ to be specified through input variables (w) (e.g., the geometric and functional characteristics of the road network) and parameters to be calibrated (supply parameters α, and demand parameters β).

The transparency of the model, and, therefore, of the plan or SG, depends on the explicitation of the following two levels:

• Explicit formulation of models and functions with their variables (or attributes);
• Indication of the values of the parameters used in the functions.

Only the explicit description of both levels (I and II) makes the representation of the phenomenon of mobility in the plan and in the SG transparent and reproducible. In the existing and available literature (see Section 1), neither the functions with variables nor the parameters are present. It should be noted that existing works referring to outdoor environments mostly present “black boxes”, i.e., they do not provide the explicit formulation of functions with variables (I) and parameters (II) of road mobility models in evacuation conditions in an urban environment.

The aforementioned studies [11,12] also report black-box SGs for all cases because they do not meet the transparency levels defined above. Based on these considerations, it is not possible to perform benchmark analyses given the lack of transparent SGs.

Based on the above considerations, to simulate an emergency scenario defined by the variables O and V for the purpose of designing a plan, it is possible to use TSMs (I) with appropriate parameters (II). The clarification of the two levels makes the representation of mobility in evacuation conditions transparent. These models, in addition to being used for planning, can be included within an SG platform; it is, therefore, possible to estimate the performance of the system in terms of exposure (E). E can be measured, for example, in terms of evacuation route times, orientation times to identify and choose routes, or the number of people leaving the system in a given time, which depend on link and route flows.

The equilibrium conditions for the ordinary state of the system are studied using deterministic (DUE, deterministic user equilibrium) or stochastic (stochastic user equilibrium) TSMs. The models for ordinary conditions can be validly adapted to evacuation conditions, assuming that the road network structure remains unchanged.

It should be noted that in “extreme” emergency situations, such as those caused by the collapse of the Twin Towers or Hurricane Katrina, transport services change significantly compared to normal conditions. This results in a change in the topology of the network (Δ) and in the number of routes available to reach safe destinations. The number of routes tends to decrease until it becomes zero in extreme conditions. It is, therefore, necessary to study how users perceive the new structure and are able to orient their travel choices in order to save themselves. As mentioned, the limiting condition for extreme situations is when only one route is available to reach a safe place. In this case, although travel time increases due to the greater volume of traffic, it does not influence the route choice (Equations (6) and (7)). Similarly, if there are only two routes, the travel time will be balanced between the two, almost identical. In this case, it is, therefore, possible to use dynamic network loading (DNL) or stochastic network loading (SNL) models.

In the total evacuation time required to move between the dangerous and shelter areas, the time taken by the individual user to identify the route to follow (orientation time) is more important than the travel time. The SG can help improve the route learning process, reducing the orientation time and, therefore, the total evacuation time, as well as the exposure level.

Table 2. Comparison between models for ordinary and emergency conditions.

System Status	Ordinary (Moderate Risk)	Extraordinary (Extreme Risk)
Number of paths in the choice set	Several (n > 2)	Limited (n ≤ 2)
Interaction models (TSM)	DUE/SUE	DNL/SNL

summarises the above, with the differences in choice set and modelling approach in the case of ordinary conditions, or moderate risk, where the network remains substantially unchanged, and extreme emergencies, where many road links are unavailable.

2.3. Exposure Reduction by Plan and Game

Exposure assessment requires the construction of transport system models that simulate the movement of people from the disaster area to shelter locations. This depends on the development of integrated TRA and TSM systems, designed according to the general characteristics recalled in Section 1.

Digital platforms support the design and implementation of SGs to experiment with real transport systems and track user’s choices under evacuation conditions.

Let S denote the generic accident scenario to be simulated. The event occurrence and the system vulnerability to the S effects are fixed and indicated with O^S and V^S. Under these conditions, the values of the variables representing exposure can be quantified with the indicator E^S.

In the formulation of Equation (1), variable E is expressed for individual risk as a probability, and its value varies from 0 to 1. In a practical application for measuring E variable, a proxy measure can be used. For example, an exposure indicator may be the total time required to evacuate the people present in the area, or the time to evacuate the last user. It is, therefore, necessary to construct a set of models (TSMs) to simulate the mobility under evacuation conditions and, thus, to estimate the exposure variable. These models allow for the quantification of variables that estimate the evacuation travel’s modes and times for each scenario S.

Through the platform that supports the realization of the SG_i, it is possible to simulate different scenarios S, and, for each one, we estimate the exposure value, E^S_i. These scenarios include (a) the scenario characterized by the absence of an emergency plan (P_not), for which the exposure value is denoted by E^S_not, and (b) the optimal plan scenario (P_min), for which the exposure value is denoted by E^S_min. With these notations, Equation (2) can be rewritten as follows:

E^S_min ≤ E^S_i ≤ E^S_not

(8)

The generic SG_i produces an exposure value (E^S_i) that is lower than the value E^S_not. Successive exercises contribute to exposure reduction towards the minimum value E^S_min estimated by the plan. The results of Equation (8) can be inserted in Equation (2) considering that O and V are fixed.

With this value, it is possible to calculate the variation between the realization from the condition characterized by the absence of the plan, as follows:

ΔE^S_i = E^S _i − E^S_not

(9)

2.4. Transport Modeling Evolution with e-ICT

In recent years, some ICTs have become available with an advanced level of maturity in different fields. Some of these are able to collect information and data for feeding transport modeling by producing an evolution of the analytical formulations of TSMs. However, in the transport sector, technological evolution continues. For this reason, in this paper, these are indicated as emerging ICT, or e-ICT. Each of these tools represents a possible option that can be activated in the proposed SG’s framework, including the following:

• Internet of Things (IoT), including a set of sensors that provide information and quantitative data diffused with the Internet network; these tools enable mutual connections between different systems;
• Blockchain (BC) or Internet of values, including a set of transactions linked in individual chains by an Internet network; this technology synchronizes data exchanges between central and distributed organizations of connected servers;
• Big data (BD) include a set of data characterized by volume, variety, and velocity, based on multidimensional matrices;
• Artificial intelligence (AI), including a set of data and algorithms, identifies alternative solutions to a defined problem and related data in an autonomous way;
• Digital twins (DTs) or a system of models and algorithms for simulating a real system; this technology allows the representation of a real system. DT technology plays a relevant role in the proposed framework for supporting serious games for representing emergency situations in a virtual context.

Each set of quantitative data provided by the recalled technology represents an input for the emergency simulations developed in the SG. The level of detail and accuracy of data and information provided by one or more e-ICT tools influences the building of transport modeling in terms of the specification of functions, calibration of parameters, and validation of the obtained simulations.

The TSMs [24] presented in Section 2.2 can be developed by incorporating the formalization of e-ICT into the demand components ([25,28,38,39]).

3. The Proposed Framework

The realisation of a Serious Game (SG) simulating the performances of a road network in evacuation conditions for scenario S requires the specification of the components described in Section 2. All components are integrated within a framework (Section 3.1) including: TRA and TSM core’s models that simulate transport systems in evacuation conditions and measuring exposure (Section 3.2); the physical devices that allow users and planners to interact inside the digital platform (Section 3.3); the e-ICT tools that represent the technologies available on the urban context (Section 3.4). Concerning the information and data provided by e-ICT tools, it is possible to upgrade the core’s models (Section 3.5) for supporting learning function, designing and implementing a prototype test (Section 3.6) and analyzing the obtained data (Section 3.7).

3.1. Main Components and Their Interactions

The formulations introduced in Section 2 require the development of a digital platform that reproduces reality as a digital twin [40,41,42]. The main aim of the platform is to simulate the quantitative effects produced by the generic evacuation plan P_i,, considering the generic user in a transparent and reproducible way.

The platform integrates the following core components (Section 3.2):

• Risk functions defined in TRA, where occurrence (O) and vulnerability (V) are exogenously defined for different emergency scenario events (S);
• Transport modeling defined in TSM for estimating the exposure component (E) in relation to different planned configurations or advanced models that consider the presence of e-ICT; this means specifying the functions, characteristics of the network, and supply and demand parameters.

Players utilize the platform via the physical interface, which comprises one or more devices, with varying levels of complexity (see Section 3.3). Each device is associated with a software or model function of the core models that represents evacuation conditions.

The platform potentialities can be enhanced with one or more elements of the relevant e-ICT tools (Section 3.4).

The generic emergency plan P_i specified by a planner can be experimented upon, proposing different alternative scenarios to the players.

In a basic experiment, the simulation consists of applying calibrated models that reproduce the emergency conditions and record the behaviour of the involved player. This is analogous to a digital twin, which replicates a planned configuration of the transport system for the emergency scenario S, for instance, in terms of evacuation routes to reach safe areas.

The outputs of the core models allow planners and analysts to measure exposure reductions in relation to the experimentation of P_i in the simulated context of the SG.

Successive SG_i experimentations of emergency plans (P_i) allow analysts and planners to quantify a priori potential effects in terms of risk reduction.

Figure 4 depicts the main component of the proposed framework and the logical flows.

3.2. Core Models

The core of the proposed framework is constituted by the models described in Section 3, including the following:

• Models derived from the TRA for estimating the risk related to different emergency scenarios (S) in terms of occurrence (O), vulnerability (V), and exposure (E); these models provide a risk function for estimating individual and societal risk indicators;
• Models belonging to TSM that reproduce supply and demand components and their interactions in the evacuation conditions; these models provide the engine for the calculation of the main transport system performances in emergency conditions (evacuation times, flows, evacuated people); this can be implemented by adopting the specifications and calibrated parameters of models available in the literature.

SG_i has to reproduce the generic risk scenario (S) by specifying Equation (1). In particular, it is necessary to define a set of functions according to TRA for estimating the following:

• The occurrence’s probability by considering the risk source’ event (i), for instance, a road accident involving a tank transporting dangerous goods; this function can be exogenous with respect to the core;
• The vulnerability’s probability by evaluating the potential consequences produced by each type of event (i) in terms of physical effects (e.g., explosion, fire, toxic clouds) and their extension in space and time and the relative resistance level; for instance the physical and geometrical characteristics of an explosive cloud, produced by a BLEVE (boiling liquid expanding vapor explosion) after a road accident involving dangerous goods [43]; this function can be exogenous with respect to the core;
• The estimation of the exposure component requires the implementation of transport modeling functions represented with Equations (3)–(5) for the transport supply component and Equations (6) and (7) for the transport demand component.

The implementation of TRA and TSMs enables analysts to simulate the transport system in emergency scenarios. The main aim of the integrated models is to obtain a quantitative assessment of potential effects produced by the actions delineated in the generic evacuation plan, P_i. The plan ensures the proper identification of the shelter’s areas and evacuation routes. These routes can be used by users to escape dangerous areas and reach the shelter. TRA and TSM core models for simulating the SG_i enable evaluators to estimate quantitative indicators of exposure variations between the pre- and post-SG_i realizations.

3.3. Upgrading the Modeling Framework

Serious games can take place in the real world, simulating emergency conditions in an urban area equipped with e-ICT technology. This is achievable in a virtual reality context where it is possible to emulate the presence of one or more e-ICT systems that provide real-time information during the emergency simulation, advancing the TSM. e-ICT integrates with the SG exercise that replicates the planned evacuation procedure, aimed at relocating individuals from areas affected by a disaster to safe zones. The incremental use of e-ICT enables analysts to obtain an ex ante evaluation of the effectiveness of the information provided to users.

Mobility during evacuation conditions results from the interaction between transport supply and demand, which evolves due to the development of the road transport system and user choices. TSMs simplify these choices by representing the quantitative characteristics of supply through a set of attributes and suitably calibrated parameters [44]. The use of e-ICTs can support the process of building TSMs through the data and information collected in databases. In this context, it is important to recall the role of big data, defined according to its volume, speed, and variety. For instance, a driving simulator represents a system that can simulate players’ travel choices under different driving conditions, including evacuation conditions.

A digital platform can create a virtual reality of an emergency situation, related as closely as possible to a potential reality. Furthermore, tracking technologies, including smartphones and connected vehicles, can aid in recording behaviour and travel choices. It is essential to maintain high levels of internal and external validity for these new technologies and their respective applications [45].

The simulation of the urban area is carried out with the core models and, thus, with the supply (3–5) and demand (6–7) equations. The presence of e-ICT in the urban area is simulated using Equations (8) and (9).

3.4. Technologies for Supporting the Evacuation Serious Game

The proposed framework enables the implementation of core models specified in Section 3.2. This requires defining the physical interface, which involves identifying a set of devices used by a generic user during the serious game.

Concerning the type of simulation object of the serious game and its requirements, it is necessary to define the characteristics of the physical interface and its components or devices. Different categories of devices include the following:

• Basic devices, including common interface tools (PC, smartphone) and related peripherals (mouse, keyboard, monitor, etc.), used for teaching purposes and characterized by relative ease of use and adoption (Section 3.4.1);
• Control devices, including common tools (PC) and peripherals (e.g., mouse, touchscreens, etc.) used for overseeing the entire SG process;
• Integrated devices, including advanced interface tools, from driving simulators used for entertainment to the more complex simulators used for experimental purposes (e.g., navigation on a spaceship) (Section 3.4.2).

Integrated devices exhibit a higher level of complexity compared to basic devices, thereby facilitating planners and analysts (referred to as supervisors) in conducting serious games within a more realistic framework through the use of control devices. These supervisors employ control devices to manage and supervise the comprehensive development of serious games.

3.4.1. Basic Devices

Basic devices belong to the simplest class of tools that enable users to interact with the digital platform used for playing a serious game. These tools are characterized by their extensive market penetration, low acquisition costs, and user-friendly design, making them accessible to a wide range of users. At the same time, these devices exhibit a diminished capacity to replicate authentic phenomena. The common basic devices include the following:

• Mobile tools (smartphones, tablets) are the most widely available category. The advantage of these is that they are accessible, making it likely that each participant has them independently during the simulation. They are useful for gamified apps and also for lighter simulations that do not require high computing power [46];
• PCs and monitors are widely available on the market with different computational potential (e.g., CPU cores) and graphical representation (e.g., computer graphics); these characteristics allow the execution of simulations with a higher level of complexity with their computational capacity and suitable software; with mobile tools, PCs facilitate dissemination processes [47];
• Virtual or augmented reality (VR/AR) tools that increase the perception of the surrounding environment and generate immersive environments that aid players in making choices more similar to those in real situations [48]; basic VR/AR devices allow for the highest level of immersion and realism in simulations; however, they also have the highest purchase and installation costs, in addition to the fact that they are not technologies with widespread use, like smartphones and PCs [49];
• Supervision consoles for managing SGs and observing players [30,50].

Basic devices can be combined in a digital platform to create the integrated devices described in the next subsection.

3.4.2. Integrated Devices

Integrated devices enable a greater level of immersion and, consequently, more realistic simulations, due to the combination of increasingly complex hardware and software. However, their acquisition demands a larger quantity of financial and human resources compared to basic devices. Among the integrated devices, the following components can be identified:

• Means transport simulators (railway, air, etc.) including common driving simulators that represent vehicle–driver–external environment interactions with different levels of sophistication [51]; this research focuses on vehicle–external environment interactions. In their simplest configuration, these devices consist of a combination of hardware (usually a PC) and an external peripheral device for driving simulation in relation to the simulated transport mode (e.g., a steering wheel or joystick);
• A combination of driving simulators and VR/AR tools for realizing an SG-based evacuation training system in a gamification context [52]. These devices allow the elements of the previously defined class to be combined with AR/VR devices;
• Tools to reproduce an urban reality through a set of equipment; an example is the Cave Automatic Virtual Environment (CAVE) typology, including a set of projectors or screens in the walls of the room [53]; in recent applications, serious games have used immersive virtual reality (IVR) to increase knowledge about evacuation procedures and support behavior assessment [12].

Integrated devices are utilized in order to simulate the performance of road transport systems in ordinary conditions [44,54].

3.4.3. Control Devices

The management and control of the digital platform that implements the proposed frameworks require the equipment of a set of devices that allow planners to define the characteristics of emergency scenarios by specifying variables and parameters and which allow analysts to observe the system and collect data and information about transport systems in an evacuation condition. These devices make up control rooms with different levels of complexity and increasing sophistication concerning the level of detail needed to represent the real emergency situations [55].

3.4.4. Assembly of Various Devices

Figure 5 illustrates basic and integrated devices according to their complexity and immersiveness. The increased complexity implies higher installation, investment, and maintenance costs but a greater level of representation of real phenomena. Integrated devices are generally exploited in professional training laboratories (e.g., aeronautics) and not simply for wide-ranging educational purposes.

The literature does not present works that combine TRA, TSM, and e-ICT to support the collection of data and information related to transport system experiments along with transparent and readable models. To the authors’ knowledge, there are no integrated devices for representing mobility under evacuation conditions, specifically regarding the user’s travel choices.

It is important to extend virtual reality driving simulators from ordinary to extraordinary conditions. One of the main difficulties in conducting experiments is reproducing users’ behaviors and their travel choices in evacuation scenarios. Digital tracking technologies that produce spatiotemporal data, such as smartphones, smart cards, and connected vehicles, facilitate the real-time observation of actions and interactions between road users.

The documented experiences should be adapted to emergency situations in which it is necessary to test the decisions, scenarios, and system configurations envisaged in evacuation plans. This concerns information on the components of the transport system and the options for travel alternatives (e.g., routes) to be followed to reach safe areas.

3.5. Functional Paths

The proposed framework can be used to simulate evacuation procedures activated for emergency situations and following different functional paths.

In this context, it is necessary to distinguish between evacuations from buildings or other enclosed spaces (“indoor evacuations”) and evacuations from open spaces (“outdoor evacuations”), for instance, a road evacuation by car. The generic exercise SG_i reproduces mobility under evacuation conditions as defined in a plan (see, for instance, [56]).

A serious game (SG) simulates a single emergency scenario that implies the activation of an evacuation procedure with different actors, namely users, analysts, and planners.

A generic user interacts with the digital platform for testing its capability to reach a safe area, following the instructions of the evacuation plan. In this case, occurrence and vulnerability assume determinate and constant values. The risk variations concern the exposure components that can be estimated by applying TSMs. The calculation of key exposure indicators can be derived from the simulation of mobility during evacuation conditions. The model specifications, attributes, and relative parameters are fixed. The measurement of a user’s capability can be performed by an interaction of the user with the gaming platform. The measure is, for instance, the evacuation time, a quantity that allows the analyst to evaluate potential risk reduction produced by SG_i. This time can be measured for successive repetitions of the same game by the same user.

A transport network defined by a planner is provided to the user by an analyst. With the SG, the user experiments with their evacuation capability in relation to the transport network. During the SG, the analyst observes the user’s performance.

The correct use of digital platforms improves the understanding and recording of mobility choices even when testing hypothetical scenarios involving the transport system. To this end, it is necessary to design specific RP/SP surveys supported by e-ICTs.

Figure 6 specifies a functional tree for realizing a serious game for simulating a road evacuation in an outdoor environment, with the relative models shown.

3.6. Prototype Test Design

The objective of an SG is to increase the individual’s level of preparedness for an emergency condition involving evacuation. The individual’s ability to know and implement the planned evacuation procedure must be improved.

The improvement can be tested and evaluated on a type of urban road network common to many cities around the world. The test was conducted on a portion of Manhattan that has a regular grid layout. Manhattan’s layout is one of the most widespread urban designs in the world, with reference to Roman, colonial, modern, and rational cities [57,58] (see, for example, Barcelona, Turin, Chicago, and Kyoto). Although there are no official statistics, this layout accounts for more than 50% of cities worldwide or their main portions. The layouts of other cities are partly radial and partly non-geometrically classifiable as irregular. Recent studies have analyzed urban grids and their efficiency with respect to specific functions [59,60], including mobility [61]. A Manhattan-type road network has a well-defined structure (Figure 7a).

Although simplified, the advantage of the prototype test using a Manhattan-type network is its transparency and replicability. Other types of urban structures, in the context of “extreme situations” as defined above, i.e., with a reduction in the routes available to reach safe areas, can easily be traced back to the scheme proposed in the test.

The studied network and the others (radial and irregular) can be formalized using the equations defined in Section 3.2, and the risk scenarios can be simulated in order to prepare risk reduction plans.

A prototypal experiment was organized in order to perform the initial validation of the proposed framework. The experiment consisted of playing a serious game (SG) aimed at testing the learning of a user who had to move around a Manhattan-type road network in an emergency situation with an “extreme status”.

The test road network used consisted of 18 avenues and 18 streets under normal conditions.

The whole test aimed to isolate and study only the learning effect and, therefore, measured the time taken by users to orientate themselves and identify the correct path. To this end, some parameters were restricted. Restricting these parameters, without removing generality, allowed the analyst to measure the benefits produced by the creation of an SG structured with replicable tests.

The transportation network and the load of flows on links and paths can be represented analytically with Equations (3)–(5) and the related vectors and matrices. In particular, the scenario under ordinary conditions (Figure 7b) is represented by the following components:

• The vector c or link costs;
• The vector g or path costs;
• The link-path incidence matrix Δ^curr that corresponds to the road network topology.

The emergency scenario (S) has a fixed level of occurrence (O^S) and vulnerability (V^S). With regard to the O^S component, it is assumed that a natural disaster (e.g., an extreme flood) or a man-made disaster (e.g., an extreme widespread terrorist attacks) will occur, causing the interruption of certain sections of the road network.

With regard to the vs. component, it is assumed that in the city, due to the disastrous event, several roads are blocked for various reasons, such as fallen trees, local fires caused by gas leaks, road subsidence, flooding of the road surface due to breaks in the hydraulic sanitary system, and cars displaced by a wave of water. Consider what happened in New Orleans following Hurricane Katrina in 2005 as a reference point.

Similarly, one can think of an extreme widespread terrorist event with different attacks that produce explosions in various city streets, thus causing the closure of the relative arc; think of the attacks in Madrid on 11 March 2004, or the attacks in London on 7 July 2005.

Each of these conditions produces a variation in the topology of the graph and, therefore, in the routes available for carrying out an evacuation procedure (Figure 7c). The network topology for scenario S is represented by the matrix Δ^S. It is assumed that, in order to reach a safe destination in an emergency situation, due to all the interruptions caused by the event, only one route is available, and this is not known to the general user who must use it in evacuation conditions.

It is also assumed that congestion does not affect the performance of the arcs and paths on the road network, as only one path is available.

The experiment simulates the operation of a platform organized in accordance with the framework presented in Section 3. In particular, a single function of the platform is tested. The function provides the user with a map of the road network on which to replan their journey from a starting point to a safe destination. The map shows which arcs are interrupted and which are available. Through a digital platform developed in the laboratory, the user must interact with the map to identify the only route available on the road network that connects the origin and safe destination. The network available after the event is partially represented in Figure 7c and represented by the matrix Δ^S. It is also assumed that congestion does not affect the performance of arches and routes on the road network, as only one route is available.

The ultimate goal of the SG is to reduce the user’s evacuation time, as defined in Equations (8) and (9). The experiment aims to demonstrate and measure the user’s learning speed in identifying the only available route. The successive values of E^S_i are calculated.

In this way, it is possible to verify whether, given the generic user, the value is reduced with each experiment and by how much. The users of the SG may initially be security operators, such as firefighters, police, ambulance drivers, and bus drivers involved in evacuations. In a second phase, the SG will be extended to all citizens who want to train on the prepared plan, with real scenarios. As mentioned above, the events that are simulated in this way range from a widespread terrorist attack to a serious natural disaster.

Users have access to a touch screen where they can view the road network and select the route in terms of a sequence of arcs on the graph.

Through the web platform and a touch screen device, users must draw the sequence of arcs that identifies the only route to follow in an emergency to reach a safe destination, starting from a node in the network.

The simulation is aimed at measuring a user’s ability to learn, through the serious game, the unique evacuation path to reach the safe place. The realization of subsequent exercises (SG_i) is aimed at increasing the user’s knowledge of the route to be followed in the shortest possible time.

For each game session, it is necessary to measure the time taken by the player to identify the correct sequence of road links. This time is a relevant variable because it represents an indicator of the learning capacity of the altered road network following scenario S. The platforms and their e-ICT components enable the automatic measurement and storage of the variable for successive tests performed by the same user. The reduction in time is a proxy for the reduction in exposure that would be achieved following each game session (ΔE^S_i).

Assuming that the plan has been developed and that it has identified the arcs that are interrupted by scenario S, there are two conditions simulated by the test, as follows:

• Initial, E_i = E^S_initial; if the user is unfamiliar with the plan in a real emergency situation, they will take the same amount of time to find the shortest route as in the simulation;
• Final, E_i = E^S_final; if the user is familiar with the plan and has the opportunity to practice using a platform, such as the one described above in real emergency conditions, they will use the last recorded time to identify the shortest route.

It is assumed that congestion does not affect the performance of the arcs; therefore, the cost functions presented in Equation (5) are constant. Introducing further complications, e.g., the presence of congestion, implies the need to consider non-linearities in the TSMs. This precludes the possibility of verifying in isolation the reduction in exposure (ΔE^S_i) as a result of increased user preparedness.

3.7. Data Analysis

In this paper, the prototypal experimentation involved a randomly drawn sample of 11 users. The sample consisted mainly of students aged between 20 and 22 (54%). Others in the sample were aged between 30 and 55 (27%), while 19% were users aged over 56. About one-third of the sample consisted of women. Each user had to identify the correct sequence of road links that constituted the unique available path represented in Figure 7c for 10 successive trials. The time interval between two successive tests was set at more than six hours. During each game session, the analyst measured the time taken by the user to conclude the game. At the conclusion of each session, the analyst meticulously recorded the time spent by each user to complete the trial. The sample was randomly drawn.

Figure 8 describes the times employed by each user for individuating the path in each trial (orientation time). Although the number of participants was limited, the experiment lasted a week in order to comply with the minimum time interval of six hours, providing all users interviewed with the same trend in terms of increased knowledge and, therefore, reduced time. The identical trend recorded among all users allow us to consider the type of function to be identical for any other samples that may be tested in the future.

A more in-depth statistical analysis of the experimental results was carried out (see

Table 3. Statistical functions for fitting data from the surveys.

Analytical Form	Function	R2
Linear	$\mathrm{t}=-2.83\mathrm{n}+34$	0.39
Polynomial	$\mathrm{t}=0.49{\mathrm{n}}^2+44.9$	0.46
Logarithm	$\mathrm{t}=13.1\ln \left(\mathrm{n}\right)+38.2$	0.48
Power	$\mathrm{t}=38.1\mathrm{n}-0.61$	0.51

). Figure 8 shows the power function that best fits the collected data (R² = 0.51). The decreasing trend in resolution times represents an increase in the learning level of the individual user. The first tests show that the experiment facilitates the learning process of the user concerning the possibilities offered by the urban road network to reach a safe destination.

The prototypal experiment assumes that the time taken by the user in each trial to individuate the safe path represents the user’s behaviour in an exercise evacuation procedure and their performance.

The maximum initial value E^S_initial among all users was 80 s and the minimum value was 20 s, with an average value of 43 s.

The maximum final value E^S_final among all users was 15 s and the minimum value was 4 s, while the average value was 9 s.

Considering the average values calculated across the entire sample, the difference between E^S_initial and E^S_initial was 34 s. The percentage reduction compared to the initial average value was, therefore, 79%. Note that this time is independent of the values of parameters α and β, considering that only one road route remains available during evacuation conditions.

Figure 9 provides a representative aggregate overview of the variability within the same test and between different tests. The average and median values of the times recorded in each test are shown. In the first three tests, there is high variability in times within the same test. In subsequent tests, the variability decreases and the times converge. This shows that with a limited number of tests, there is an improvement in the learning process.

The mere availability of a good plan is necessary but insufficient condition for risk reduction. The results show the need to test a plan in order to contribute effectively to reducing exposure through evacuation. Testing, although conducted on a limited sample of users, indicates that it is possible (and, therefore, socially necessary) to conduct a series of a few tests (three or four tests) in order to achieve a reduction in orientation time of more than 50% (see

Table 4. Trial numbers and time reduction.

	Time Reduction by the First Trial
Analytical Form	25%	50%	75%
Linear	4	7	9
Polynomial	2	4	8
Logarithm	2	4	8
Power	2	3	10

).

4. Discussion

The findings of the prototype test demonstrate the merits of implementing emergency drills, and, in the specific case of this study, serious games. The results of the experimentation tests show that, through the implementation of the subsequent trials, the resolution time has a decreasing trend. This confirms that the user identifies the solution more quickly and, thus, increases their knowledge of the altered graph topology. In a real-world context, this corresponds to an increase in the level of preparedness to relearn the alternative routes proposed by the plan to reach safe areas more quickly. Conducted exercises increase the user’s level of network knowledge, which, in a real emergency, can easily be put into practice.

The findings indicate that repeated experimentation within the same paradigm consistently yields optimal outcomes with regard to both temporal efficiency and exposure. This work contributes to the existing body of literature by addressing a significant gap in research, namely the absence of tests that measure the reduction in evacuation time in a city and, thus, exposure (E) through a reproducible game.

The simplification adopted in the case study allows for the isolation of the effects on user learning with reference to the only alternative route available in the event of an emergency. Even in congested conditions, the choice would still consist of a single alternative. Therefore, the experiment represents a basic level that already offers a wealth of information about the value of SGs, even in comparison with other types of exercises.

The simplified schematization of the test problem with respect to the type of network attempts to reproduce the road network after a very serious disaster. As discussed above, in the event of very serious events, it is necessary to provide tools in advance that can facilitate the orientation and identification of users in an emergency. Given the limited number of routes, the problem of orientation is crucial, together with that of travel time, which, in the extreme case of only one route being available, makes the choice independent of congestion.

Starting from a basic level of the framework, it is possible to implement more complex framework applications. However, this work started from an initial condition. It is necessary to assess the impact of using VR/AR technologies. The prototypal experiments require more insights to better represent evacuation processes in a transport system in an urban area. In the proposed example, there are no influences on congestion-related effects, and there is no availability of e-ICT.

Despite these limitations, the first results are promising and require more development in further work concerning effective disaster risk management. The results show the importance of performing preparedness activities in order to improve the effectiveness of emergency planning. This is possible through the design and implementation of a cycle of exercises with an increasing level of complexity, aimed at increasing the ability of those involved to respond to emergencies, designing the best plan (planners), and following the designed evacuation plan (users). This leads to a reduction in the gap between the decisions made during the mitigation process, reported in the plan, and the decisions actually put into practice during a real event. Serious games, inside the discussion-based exercise activities, contribute to reducing this gap, and today, with the expansion of digital platforms, can be very important. The availability of emerging ICT technologies together with TRA and advanced TSMs allow for the evolution of classic games towards serious games, which, according to the literature, combine experience, multimedia delivery, and entertainment.

With regard to the exposure element, serious games have the capacity to replicate a transport system in the context of an evacuation scenario, thereby simulating user travel decisions. In this context, it is necessary to specify and calibrate TRA and TSM functions that simulate the evacuation procedure accurately and as close to reality as possible. By considering a specific disaster scenario (S), consolidated TRA functions can be adopted for estimating occurrence and vulnerability. The measurement of exposure component implies the specification, calibration, and validation of static or dynamic TSMs that can be integrated with e-ICT. It is necessary to reproduce the user’s behaviour by updating utilities and travel choices in an evacuation condition. This requires the use of devices defined by increasing levels of complexity and immersiveness. SGs allow users to experience an emergency situation and experiment with the effects of their choices in a simulated environment where the variables and parameters of the models have been calibrated and remain fixed. The data collected during the SG allow planners and analysts to evaluate risk reduction, measuring distance from the ideal conditions defined by the best evacuation plan (P_min). In this case, the variables and parameters of the models are subjected to modifications with the aim of reproducing scenarios that can reduce exposure and, therefore, overall risk. The aim is to increase users’ preparedness about evacuation planned procedures to be implemented during a disaster event with the aim of reducing risk.

5. Conclusive Remarks and Future Perspectives

Limiting damage produced by disasters is currently a priority challenge at the international level, as specified by Agenda 2030. This implies the need to design and adopt emergency planning for defining a priori actions to be implemented when a disaster occurs.

The emergency planning process needs to be supported by exercises that simulate the chosen actions for reducing risk and their components. The main aim of a generic exercise regards the reduction in the gap between the decisions taken in the mitigation phase and their implementation in reality, which can constitute the contents of the preparedness phase. For instance, the prototype experiment of this paper shows that a reduction in solution time of more than 79% from the first trial is achieved by performing successive trials of the same game.

Traditional games are already a consolidated best practice in some parts of the world. Moreover, there are already experiences on an international scale using computer games aimed at simulating certain emergencies (e.g., the UN for tsunamis, see [62]). This paper emphasizes the importance of SGs for increasing situation awareness and measuring risk reduction. SGs represent the most advanced level of discussion-based exercises and, in some cases, could be more effective and less expensive than operation-based exercises. The main advantages of SGs are their reproducibility and replicability over time at a low cost, given that they can be performed in a digital and virtual environment. The experiment conducted and the results obtained, despite the simplification used, show that it is possible to measure the benefits deriving from the design and implementation of an SG. The proposed framework could improve existing disaster drill protocols in order to reduce the gap between planning and real emergencies. Among other things, the results of the experiment show that just a few successive trials are enough to improve users’ learning levels. This implies the necessity to (1) comprehend how digital platforms can be used to support exercises for increasing preparedness and collecting data in the field; (2) provide more insights about the potential contribution of e-ICT for improving the accuracy of TSM modeling approaches for simulating mobility in evacuation conditions and then estimating exposure. This paper underlined the necessity of combining e-ICT with TSMs for the simulation of mobility under evacuation conditions. This must be realized in a transparent way and not using “black boxes”. On the one hand, it further reduces the gap between the mitigation and preparedness phases, as is the case with any emergency exercise. On the other hand, the results became inputs of a modeling framework for transport network design in a transparent way. Some technologies have been developed for realizing digital twins of transport systems in ordinary conditions and driving simulators are used for collecting data about road user’s behaviors in an ordinary condition. This paper gives the framework to extend the potentialities offered by current e-ICT with TSM developed for evacuation conditions, extending simulator functionalities. This makes it possible to simulate planned emergency procedures in a virtual environment that reproduces real urban contexts characterized by the presence of different availability levels of e-ICT. A digital platform, designed with the above characteristics, would allow policymakers and emergency managers to experiment with different physical and technological configurations of the same city, under different risk conditions represented in a virtual environment, in order to increase the preparedness level and capabilities of communities.

This paper underlines some limitations present in the literature in relation to the deep specifications’ definition of users’ requirements, including functional and physical architecture for designing and releasing a digital platform for supporting an evacuation SG for a general urban area. This problem requires more insights in relation to the implications deriving from the combination of TSMs and e-ICT inside TRA. The platform can be developed in a prototypal way using current technologies available on the market.

The proposed framework needs further specifications to improve the design and implementation of SGs. It can be expanded upon by introducing increased levels of complexity related to their components.

The prototypal experiment has to be extended by increasing the number of participants and the level of complexity and immersiveness, followed by the use of more complex digital tools (e.g., AR/VR). The serious game can be performed by multiple users and by considering the reciprocal influences between the participants. In future extensions of this research, it will be necessary to test the proposed methodology on additional types of road networks with more complicated structures than the Manhattan style. This is necessary to assess whether the learning process can be influenced by the topology of the network.

Future developments in this research will make it possible to easily extend the analyses to radial or irregular road networks, using graphs that, under extreme conditions, are as complex as the test graph. Further studies will make it possible to verify the validity of the learning functions identified. This work may be useful for comparison with other research that meets the transparency levels defined above, in order to confirm or reject the results obtained.

Another element of complexity regards the simulation of the effects produced by choices of planners and analysts. Furthermore, the complexity concerns software and hardware included in the digital platform, with reference to the basic and integrated devices. According to the research’s objectives, the platform can be constituted by open-source software, like GIS, which is useful for territorial representations. The most advanced level regards the interactions between planners, analysts, and multiple players in a virtual geospatial environment. Each level of complexity implies advancements in data elaborations and core model development performed by analysts. The final aim is to enhance the capability of the serious game to experiment with combinations of risk reduction strategies, related to one or more risk components (O, V, and E).

In further developments of this research, it will be necessary to adopt an interdisciplinary approach to increase the knowledge of the phenomenon of mobility in evacuation conditions. This implies the necessity to consider the other risk components (occurrence and vulnerability). In this context, it is relevant to produce more insights into the effects produced by information provided by decisionmakers to users addressing travel choices to achieve increased levels of safety and security.