1. Introduction
During the last decade, the tourism industry significantly contributed to economic growth (Figini and Vici, 2010). Passenger ships carrying at least 12 passengers, including cruise ships and passenger ferries, make up a significant part of society's revenue. Thirty million passengers are expected to travel on cruise ships, generating over $154 billion in revenues worldwide in 2019 (Cruise Lines International Association, 2021). Conversely, traveling by sea increases the safety risk for passengers. Allianz (2021) reported 69 passenger ship losses from 2011 to 2020. In addition, see Table 1, at least 2526 people lost their lives due to incidents from 2011 to 2018.
Year | Ship name | Type | Fatalities | Reference |
---|---|---|---|---|
2011 |
|
|
|
Fundi (2018) |
2012 |
|
|
|
Vanem and Skjong (2006) |
2013 |
|
|
|
Fahcruddin et al. (2019) |
2014 |
|
|
|
Kim et al. (2016) |
2015 |
|
|
|
Baird (2018) |
2016 |
|
|
|
Christine and Bonnemains (2018) |
Total |
|
- a
-
203 passengers died, and 1326 passengers are still missing but presumably dead.
The facts mentioned pushing IMO to enhance safety at sea. The Maritime Safety Committee (MSC), which is primarily responsible for coping with all safety issues at sea, published principal safety regulations through different circulars (Circ.) (IMO, 2016). They aim to upgrade basic maritime safety standards for ships, first released by the International Convention for the Safety of Life at Sea(SOLAS) in 1914. Evacuation models have been integral to the issued regulations. Xie et al. (2020a) pinpoints evacuation as an effective action for lowering the casualty rate at sea. A ship evacuation process occurs in three successive distinct periods: (1) response, (2) evacuation, and (3) embarkation and launching period (IMO, 2016). Evacuation time is the central part of the evacuation process. It must not exceed the onset of circumstances threatening passengers’ safety. Initially, the response period starts off noticing initial notifications (e.g., alarm) until deciding to move. Then, the evacuation period starts from the moving point to an assembly station. Afterward, the launching period commences. The mustered people in the assembly stations (or embarkation stations) must abandon the ship with a ship signal to reach a safe place. If the assembly and embarkation stations are separate, there is also a travel time between the assembly and embarkation stations.
Meanwhile, evacuation factors have a critical function during the evacuation process. Various factors influence the process, including environmental, configurational, behavioral, and human. Table 2 categorizes them according to definition (Lee et al., 2003).
Aspect | Definition | Features |
---|---|---|
|
|
|
|
|
|
|
|
|
|
|
|
Fig. 1 depicts a ship evacuation process sequence considering influencing elements.
The MSC has pushed ship designers to analyze the evacuation process by putting the evacuation factors into practice. Considering the factors in modeling at an early stage of ship design can preclude any extra safety assessment later in calculating evacuation time. Specifically, it can reduce the possibility of rebuilding ships for only satisfying new safety evacuation standards. Simplified and advanced analysis are two categories of evacuation analysis (IMO, 2016). A simplified analysis is according to a deterministic method assuming passengers as nonautonomous agents. However, the latter considers passengers autonomous agents under the uncertain influence of input parameters, such as ship motion (Nasso et al., 2019). Ship designers should implement relevant corrective actions if the evacuation time exceeds the allowed time. Existing passenger ships could also carry out appropriate evacuation actions to reach the permitted evacuation duration (IMO, 2016). Given the above discussion, two research questions arise:
-
•
What is the current situation of evacuation models for human evacuation from passenger ships regarding evacuation factors, modeling approaches, and solution methods?
-
•
How do evacuation factors affect human behavior in the event of an accident?
This paper presents a review to identify the current situation and create a roadmap for future research in this area. The authors have not identified any comprehensive literature overview in this domain. This paper tries to cover this gap by reviewing, categorizing, and analyzing 105 publications released between January 1999 and August 2022. The specific review choices resulting from this sample of papers are explained in detail in Section 3.1. Before coming to that section, the authors first discuss earlier review/partial-review papers in Section 2. Research methodologies are clarified in Section 3. Detailed analyses and classifications are coming in Section 4. The current gaps analysis and future research opportunities are presented and discussed in Section 5. Finally, Section 6 contains the conclusion and directions for future research.
2. Literature review
Understanding the state of the current literature establishes a firm base for advancing knowledge and uncovering novel research areas (Pignatelli et al., 2005). Therefore, the previous review/partial review papers and IMO guidelines are listed in Table 3.
Scope | Regulatory reference | Range | Sample size | Paper |
---|---|---|---|---|
|
|
1995–2001 | 25 | Lee et al. (2003) |
|
|
2001–2008 | 20 | Yip et al. (2015) |
|
|
2002–2015 | 10 | Bucci et al. (2016) |
|
|
1973–2017 | 53 | Sarvari et al. (2018) |
|
|
1974–2018 | 57 | Stefanidis et al. (2019) |
|
|
|
23 | Nasso et al. (2019) |
|
|
|
115 | Our study |
Given Table 3, the authors have been unable to identify any comprehensive review study for human evacuation analyzing state-of-the-art research papers considering evacuation factors. Most review papers in Table 3 are limited based on the scope and period. In this area, review papers are divided into two groups. The first group of review studies is based on assessing past findings and the current situation. For instance, Sarvari et al. (2018) and Yip et al. (2015)investigated a range of publications for a specific period. The second group of review papers is according to the IMO guidelines for analyzing evacuation methods for passenger ships. For example, Bucci et al. (2016), Lee et al. (2003), and Stefanidis et al. (2019) primarily focus on examining the IMO guidelines for passenger ships.
Among all mentioned papers in Table 3, Sarvari et al. (2018) almost reviewed all relevant academic journals and conference papers on human evacuation from passenger ships. Although they analyzed the influencing evacuation factors on the evacuation process, the number of publications in their analysis is low. Furthermore, covered papers were published before 2017. The current work coincidences with Sarvari et al. (2018). The reason is about 60 percent of publications are released between 1999 and 2017. Therefore, this paper attempts to include them in the database and examine them based on the defined objectives, for example clustering the collected publications according to research type, quantitative (modeling or data collection) or qualitative (questionnaire, case study description, or evacuation software analysis). In addition, although Gwynne et al. (2003) and Galea et al. (2014a), with 61 and 31 citations until March 2022, are disregarded in the review of Sarvari et al. (2018), they are reviewed in this paper since they offer a significant contribution to the process of data collection and validation for human evacuation from the passenger ships.
Moreover, 1999 was a watershed year when the IMO issued the first circular regulations of evacuation analysis for passenger ships, so 1999 is applied as the cut-off year. Following this synthesis, this study intends to present a systematic review of the influencing factors on the human evacuation process for passenger ships and appropriately determine the modeling approaches and solution methods. At the same time, this paper looks at how emerging technologies such as digital twin (DT) and virtual reality (VR) can shift the performance of the evacuation process. Fig. 2 depicts the trend in the number of publications over the study period. Although research has been active during the first decade (1999–2009), this area has received more attention over the second decade (2010–2022).
3. Research methodology
This paper applies a four-step process to analyze the content in literature reviews. This process aligns with the qualitative content analysis methodology (Mayring and Brunner, 2007). It consists of (1) material collection, (2) descriptive analysis, (3) category selection, and (4) material evaluation steps portrayed in Fig. 3.
3.1. Material collection
The current research was carried out from September 2021 to June 2022. This paper covers peer-reviewed papers, conference papers, and doctoral and master dissertations in scientific English language journals from January 1999 to August 2022. The material collection is conducted in five stages. The stages are as follows.
-
•
Identifying keywords: they are referred to our research questions to identify the keywords. Therefore, the keywords are defined as" passenger ship (cruise, ferry, ocean liners), evacuation, emergency, human/passenger behavior."
-
•
Defining search strategy: this paper pursues a search string strategy. It combines keywords, truncation symbol (keyword root + *), and boolean operators (AND to include all search terms, OR to include alternative terms, NOT to exclude specific terms). The search is conducted in title, keywords, and abstract.
-
•
Searching in databases: the authors selected the Web of Science (WoS) database for gathering material. Likewise, the search is carried out on the Open Access Theses and Dissertations database (OATD) to identify relevant research.
-
•
Crosschecking in publishers: the collection is crosschecked by publishers to determine records' accuracy to include/exclude the intended keywords.
-
•
Reporting outputs: the selection set is first transferred to Excel sheets for data cleaning and organizing collected papers. Afterward, the database is called in Python data frames for analysis and visualizations. Pandas, NumPy, and Matplotlib libraries are employed to analyze data.
Fig. 4 demonstrates the employed search strategy for retrieving relevant studies.
The authors initially searched on WoS. The search yielded papers consisting of at least one of the keywords in the first step and a word from the root of evacua (evacua*). It produced 4300 records. After that, any paper concerning evacuation from buildings, hospitals, trains, aircraft, and stadiums is excluded. The excluding strategy stood on WoS's advance search option and the authors' inspection by reading the paper's title (if necessary, the abstract is read as well). Similarly, theses and dissertations are retrieved from the OATD database. Ultimately, 115 papers are downloaded and classified. The records are distributed as 27% from Elsevier, 24% from Springer, 6% from IEEE, 4% from MDPI, 3% from Taylor & Francis, 3% from National Taiwan Ocean University, 3% from Royal Institution of Naval Architects, 2% from OnePetro, and 2% from Hindawi. Other publishers with one publication had an overall 26% contribution. Appendix A contains the list of publishers. The records' credibility in data collection, analysis, and reporting is fulfilled by the author checking (Creswell, J.W., & Miller, 2000).
3.2. Descriptive analysis
The authors found 111 journal and conference papers, two doctoral dissertations, and two master theses- 115 research publications in total. The distribution of journals, conferences, and university publications is represented in Fig. 5. Those with more than two publications are described under the same name of journal/conference; however, others with one publication are allocated to the other categories named Others (Journal Papers/Conference Papers). They are listed in Appendix B. Appendix C is also constructed for the journal names’ abbreviations. Fig. 5 reveals a growing tendency in evacuation studies for passenger ships. Among journals, Ocean Engineering (Ocean Eng.) and Safety Science (Saf. Sci.) have the largest number of publications, with 10 and 6, respectively. They have been more active in this area. They mainly researched passenger behavior/awareness, walking speed, safety perception, and risk analysis during a human evacuation from passenger ships. Meanwhile, the Pedestrian and Evacuation Dynamics conference published more papers than others, with seven amid conferences. The released papers not only focus on data collection concerning movement and evacuation dynamics of passengers considering ship motions, but they also have worked on the simulation and modeling of human evacuation.
Besides, Journal of Marine Science and Technology (JMST), Journal of Marine Science and Technology (J Mar Sci Technol), Physica A: Statistical Mechanics and its Applications (Phys. A: Stat. Mech. Appl.), Procedia Computer Science (Procedia Comput. Sci.), Mathematical Problems in Engineering (Math. Probl. Eng.), Reliability Engineering & System Safety (Reliab. Eng. Syst), Journal of Marine Science and Engineering (J. Mar. Sci. Eng.), and Computers in Industry (Comput Ind) with 14.8% contribution attempted to reflect new insights in this research area. They tried to develop system simulation models considering passengers’ characteristics. While International Conference on Human Factorsin Ship Design and Operation, Traffic and Granular Flow (TGF), and International Conference on Virtual, Augmented, and Mixed Reality (VAMR), with a 5% impact, push the research in this area forward. They are more inclined to manifest human factors into ship design. Moreover, the University of Greenwich, University of Huddersfield, Norwegian School of Economics (NHH), and Aalto University (with a 3.5% impact) generate new knowledge in this area. They devoted considerable effort to advancing the understanding of passenger behavior during a ship evacuation. Furthermore, Springer with 40%, Elsevier with 16%, and the Royal Institution of Naval Architects with 8% have a remarkable impact on emerging the research area of human evacuation from passenger ships. They built the foundation of knowledge by conducting questionnaires, conducting onboard experiments, and simulating/modeling the human evacuation process from passenger ships. Afterward, IEEE (6%), National Taiwan Ocean University (6%), Hindawi (4%), and Taylor & Francis (2%), along with Springer (28%) and Elsevier (26%), shifted the state of research in this area and yielded fresh insights into the research by analyzing advanced evacuation methodologies and taking human behavior properties into account. Since 2019, evacuation studies have received more attention from MDPI and IOS Press databases. They accelerated development in this area by applying multidisciplinary approaches, particularly computer science, mathematics, engineering, and environmental science.
Further, it is essential to identify the subject areas of publications. The research area of each publication is placed using the WoS subject area feature. This identification can enable researchers to recognize the research area's focus and open new research topics for future research. According to Fig. 6, engineering with 38.3% is more interested among researchers, followed by 36.5% for multidisciplinary approaches. Afterward, computer science and mathematics accounted for 12.2% and 3.5%, respectively.
Moreover, those research areas with two or fewer publications are listed in the others' category (physics, environmental science, psychology, construction building technology, neuroscience, and medicine). Interdisciplinary research pays increasing attention among scholars in this research area. The reason is the presence of different evacuation factors involved in the human evacuation process. There is a need to consider all together to fulfill IMO's requirements. Integrating techniques from other disciplines, such as engineering, environmental science, oceanography, operations research, management science, etc., augment the understanding and describing human evacuation problems from passenger ships.
Next, from Fig. 7, Asia (52.2%) and Europe (44.3%) have the most significant academic contribution among others (Africa and South America have zero publications). Most publications in Asia are researched in Chinese and South Korean maritime institutions, with 39 and 12 papers, respectively. One of the solid reasons for the importance of this area for Chinese and South Korean scholars can be the sinking of Dongfang Zhi Xing and MV Sewol passenger ships with the loss of 442 and 304 passengers and crew, respectively (Baird, 2018; Kim et al., 2016). Hence, the Chinese and South Korean maritime sectors have inspired researchers and ship designers to reach safer evacuation solutions at sea. The UK (17 papers) and Norway (7 papers) have been more active and interested in Europe. Not only a disaster such as Costa Concordia and MS Scandinavian Star but also IMO's guidelines have pushed the maritime industry to rise in research and development activities in terms of human evacuation modeling. Other countries on the list have a 34.8% contribution (Japan, Greece, Germany, Poland, Taiwan, Finland, Italy, Spain, Netherlands, Sweden, Canada, USA, Australia, Croatia, and Turkey). It shows that the popularity of human evacuation problems is growing among scholars in different geographic regions. Eventually, according to the first author's affiliations, Edwin Richard Galea from the University of Greenwich with seven publications, and Xinjian Wang from Dalian Maritime University with six publications, have the most considerable contribution in this research area.
3.3. Category selection
The structural dimensions of the current research and chief topics of analysis, including detailed analytical categories, are represented in Table 4. Each category consists of different features discussed in greater detail in Section 4. The MSC scope has various study subjects, such as updating the SOLASconvention, piracy and armed robbery against ships, and cyber security. However, the current work focuses on emergency evacuation from passenger ships. The present study targets the evacuation factors listed in Table 2 to determine underlying features. Then, the collected papers are analyzed and categorized concerning the identified features. A detailed presentation of all publications in different analytical categories is described in Appendix D to H.
Analytical category | Features | Appendix |
---|---|---|
Modeling approach |
|
D |
Traffic assignment formulation |
|
E |
Model parameters |
|
F |
Hazard type |
|
G |
Solution method |
|
H |
Fig. 8 demonstrates the popularity of different modeling approaches for representing the behavior of the problem. The most significant portion of researchers prefers to apply simulation approaches for defining their model (with 52.2%). It is followed by hybrid (simulation/mathematical, simulation/experimental, simulation/questionnaire) and experimental approaches with 19.1% and 17.4%, respectively. 7% of the used methods account for the mathematical approaches. Only a minority of researchers, 4.3%, prefer to employ analytical modeling approaches. It shows the popularity of simulation techniques and increasing attention to hybrid and experimental methods.
The other analytical category is traffic assignment formulation. Karabuk and Manzour (2019) classified land-based evacuation models into an optimal system formulation and a user equilibrium formulation. Their definition is considered for ship-based models—the former attempts to offer an evacuation plan to improve the main objective (macroscopic perspective). In contrast, a user equilibrium formulation generates an evacuation plan based on the characteristics of each passenger and addresses the problem at a more granular level (microscopic perspective). For example, an evacuation model can minimize the overall evacuation time with and without considering passengers' physical condition. The former can be in the first category; however, the latter focuses on every passenger's mobility. Moreover, 64.3% of researchers address their problem from a user equilibrium perspective. It shows there is increasing attention to understanding passenger behavior in this area.
Model parameters are the third analytical category. Parameters are concerned with the model's configuration. For example, the model's settings can vary according to the ship's motion mode. Fig. 9 demonstrates how many times two parameters are assessed together in the collected papers. Thirteen factors interacted more with each other among other evacuation factors. The blue circles indicate how many times two parameters are viewed in modeling simultaneously, while the green ones indicate a parameter is solely applied. For example, ship stability is repeated ten times with passenger age. Conversely, yellow squares ascertain the gaps for future research.
The next category is the hazard type. Hazard refers to a potential source of damage to a passenger ship or people onboard. However, when the hazard happens, it can become a disaster (Shi, 2019). The most significant proportion of papers disregards considering the kind of hazard that threatens passengers' lives. In contrast, 21 publications consider fire as a hazard. Six research papers examine flooding and storm, with three for each. Although foundered (sunk and submerged) accidents with 54.4% of the total losses in the world ocean are the most frequent hazards that ships encountered from 2010 to 2020 (Allianz, 2021), only six papers addressed it in the literature. Three papers also take two hazards into account simultaneously. Also, Allianz reported wrecked (grounded): 19.6%, fire and explosion: 11.3%, machinery failure and damage: 5.8%, collision (involving vessels): 3.4%, hull damage (holed, cracks): 3%, and other causes (piracy and miscellaneous): 2.5% are other hazards. They can be other directions for future research to consider in formulating and analyzing human evacuation models.
Finally, the solution method category is analyzed. Applied solution methods are categorized based on the paper's objective. 31.3% of researchers used an evacuation tool to simulate the process. Some are based on discrete models allowing agents to occupy a discrete set of points in terms of space representation (such as MaritimeEXODUS and IMEX). In contrast, others are continuous models considering a constant sequence without interruption between different points in a defined space (such as VELOS and Pathfinder). Moreover, hybrid tools benefit from both models' properties (e.g., EVI and EvacSim). Appendix I lists available evacuation simulation tools in the literature.
The collected papers are thus evaluated and analyzed according to the features described in Table 4. The details of the analytical dimensions of the review study are thoroughly discussed in the following sections.
3.4. Material evaluation
The sample papers are cross-checked with another database, including Scopus whereby the authors verify the paper's properties, such as the research area. The aim is to improve the validity of the analysis. The author checking technique is consequently applied to control the credibility of the sample papers. After reading the abstract, they would be kept if they are consistent with the study objectives. Finally, the collection with 115 publications is established for further analysis.
4. Detailed analyses of the literature
This section gives the results of the analysis. The collection is studied according to analytical categories to determine the status in this research area. The gaps are identified, and the future research agenda is accordingly established. Although there can be an overlap in classification, this paper tries to categorize them according to the objective of each paper appropriately.
4.1. Problem classifications
There are various subjects in this area of study. Although the authors pursued particular aims, the papers can be classified into the following categories. Appendix J classifies papers according to their objectives.
4.1.1. Traffic assignment category
This subsection tries to classify the collected papers based on the traffic assignment analytical category. Papers with the aim of optimizing the overall performance of the egress system are placed in the evacuation time and route optimization subcategory. In comparison, the passenger behavior modeling subcategory pays attention to papers with user equilibrium formulation features.
4.1.1.1. Evacuation time and route optimization
Evacuation time optimization gains a significant portion of research objectives in the collected papers. All research subjects with the same subject matter are included in this classification (response time, assembly time, and embarkation time). The aim is to minimize the evacuation time considering evacuation factors. Furthermore, the route optimization module intends to provide safe evacuation routes in which the characteristics of passengers/crew distribution and the ship's layout are considered. The authors aim to determine the emergency evacuation routes available for evacuees or analyze the operational level considering congestion and counter-flow movements.
This kind of research has several advantages. For instance, total evacuation time calculation can be employed for updating the whole evacuation time in a real-time emergency response. Specifically, it can assist crew and passengers in handling the remaining time based on the available evacuation routes (Lin and Wu, 2018). Conversely, it has some shortcomings. For example, it lacks to consider passengers as conscious agents in a real-life case. Explicitly, how different aspects of passengers, such as the level of compliance, can affect the total evacuation time. Critically, this paper attempts to categorize them to represent a clear view of estimating the whole evacuation time in the presence of evacuation factors. This category consists of 48.7% of studies.
4.1.1.2. Passenger behavior understanding
31.3% of publications attempted to focus mainly on understanding passenger behavior. It is critical during evacuation as it minimizes total evacuation time and casualties in emergency maritime situations (Finiti, 2021). Many authors attempted to advance understanding of passenger behavior by finding the most significant drivers, such as ship stability (mostly considered trim and heel angle) and disaster development (most researchers considered fire), in their reaction to the emergency. Some carried out a series of evacuation trials at sea to calculate passenger gait speed under predefined emergency scenarios. In contrast, some conducted questionnaires to explore new insights with an interactive study with passengers (Deere et al., 2012; Wang et al., 2020c; Yip et al., 2015).
The main advantage of considering passenger behavior is to design an effective emergency evacuation system to ensure safety standards (Wang et al., 2020c). Likewise, passenger behavioral responses to an emergency can enhance understanding of control efforts and crowd behavior (Li et al., 2019). While understanding the various source of uncertainty in passenger behavior calls for more investigations and quantifications in this research area. Specifically, how internal and external drivers, such as stress level and ship motions, can impact the behavior. Hence, the relevant samples are categorized to reveal the importance of evacuation factors in behaving passengers during an evacuation process.
4.1.2. Solution methods
The solution method category is the next analysis classification. It consists of three subcategories: (1) description of evacuation models, (2) data collection and validation, and (3) optimization solvers.
4.1.2.1. Description of evacuation models
Another category represents the description of evacuation models (11% of studies). Parts of the collected papers described maritime evacuation models to understand the evacuation process better. Some analyzed the current evacuation models considering simplified and advanced approaches, while others tried to evaluate evacuation simulation tools (Miyazaki et al., 2004; Sun et al., 2018a). The offered category can deliver a clear view for selecting a simulation-based evacuation tool according to the models’ configuration. K V Kostas et al. (2014a) reflected the applicability of VELOS for assessing passenger and crew activities in normal and hectic conditions of evacuation operations. Also, Guarin et al. (2014) described the concept of escape and evacuation from passenger ships using the pedestrian dynamics simulation tool EVI.
Although the available simulation-based evacuation models can provide solutions, there is a need to design a real-time decision support system to track the evacuation process. It is suggested that the system can be based on a data-driven multistage optimization framework. Various real-time operation data is obtained from different agents involved in the evacuation process. They can be modeled with machine learning (ML) techniques under the uncertain development of a disaster (Roy et al., 2021).
4.1.2.2. Data collection and validation
Researchers tried to collect data through either paper-based methods, including surveys or questionnaires, interviews, or computer-based techniques such as video cameras (7% of the collected papers). Regarding questionnaire surveys, some researchers tried to analyze different points of the passengers' views during the evacuation process. They determined the impact of various factors on the evacuation process and passenger behavior. For example, Liu and Luo (2012) and Lozowicka (2021) analyzed the influence of demographic differences, including age, gender, educational level, mobility level, experience onboard, and traveling companion, on passengers’ behavior and safety awareness and perception during an emergency evacuation Ro-Ro passenger ship.
Moreover, Finiti (2021) applied two different methodologies (case study and interview) as complementary tools. They attempted to understand likely passenger behavior by analyzing the collected data from some survivors of the Costa Concordia disaster in terms of gender, age, companions, and experience. Furthermore, data related to passenger behavior under different circumstances, such as ship stability angles, play an essential role in understanding the evacuation process. Actual onboard experiments can further shift our understanding of the evacuation process. A notable example is five full-scale semi-unannounced assembly trials performed at sea under the EU Seventh Framework Programme project SAFEGUARD (IMO Fire Protection Sub-Committee, 2012). The aim was to generate passenger response time data, validation, and calibration data sets for ship-based evacuation models and establish a set of fire and trim/heel scenarios. Studies with the same subject matter fall in this category. Video-based observation is another popular method for gathering data in evacuation studies (Galea et al., 2014a; Na et al., 2019; Wang et al., 2021a).
This category can deal with relations between different data collection methods and evacuation factor data. It can help researchers find a suitable database based on their problem requirements. For instance, Deere et al. (2012) is a source of human factors data for the passenger assembly process on large passenger ships. They utilized hybrid methods, including video cameras and infrared beacons. Still, there is room for measuring biological and psychological passengers' cognitive states, such as stress levels, and how they affect the evacuation process. Moreover, sociological-based data, such as cultural diversity, can give ship managers more insight into passenger behavior (Galea et al., 2015; Zhang et al., 2017). Therefore, they are another challenge in collecting and analyzing data in this research area.
4.1.2.3. Optimization solvers
In the simplified version of evacuation analysis, the overall performance of the evacuation system is critical, whereas, in the advanced version, the egress of each human while various factors, such as hazards and ship motions, affect the behavior is the primary objective. In doing so, two types of methodologies are provided.
On the one hand, researchers use various approaches to solve the formulated evacuation problems for passenger ships. The authors have split the applied methodologies into three main categories according to the paper's objective. Firstly, many authors use simulation and mathematical tools such as MaritimeEXODUS, VELOS, CPLEX, and MATLAB to reach a solution. Secondly, some employed optimization models to harness uncertainty of the different elements of the evacuation process, such as human evacuation behavior. They include Polynomial chaos- (PC) and Monte Carlo-based (MC). For instance, Xie et al. (2020a) applied PC expansion with Gauss quadrature to quantify the uncertainty of evacuation time for passenger ships. Furthermore, Wang et al. (2013) employed an MC-based sampling method to analyze available safety egress time under ship fire (SFAT). Thirdly, researchers applied meta-heuristic algorithms for solving real-life evacuation problems (Lozowicka, 2021). For example, Ćozowicka (2010, 2005) utilized the Genetic Algorithm (GA) to find the shortest evacuation time and route. Although GA can propose a feasible structure fitted to problem parameters, there is a possibility of falling to the local optimum for this algorithm. Also, the degree of complexity is raised by considering more evacuation factors. Therefore, it is suggested to employ hybrid techniques to escape it. For example, Kaveh and Ghobadi (2020) presented a hybrid evacuation model using the graph theory and metaheuristic algorithms to find the best evacuation route under a fire situation considering human factors.
On the other hand, passengers and crew are characterized as unique individuals with distinctive personality traits and cognitive abilities. Fifty publications applied microscopic models, including social force-based, velocity-based, acceleration-based, and CA, to work out the dynamic behavior of passengers. The most considerable contribution was seen for velocity-based models with 52%—cell-based and social force-based models comprised 26% and 14%, respectively. The minor portion stands for accelerated-based models with 8%. However, there can be a potential extension to study the influence of evacuation factors, particularly dynamic conditions of the ship, on passenger behavior within microscopic-based models (IMO, 2016). In land-based evacuation path planning, Yang et al. (2022) integrated three forces, i.e., pedestrians' self-driving force, the pedestrian's interaction force, and the interaction force between pedestrians and obstacles, in the format of a social force model. Furthermore, Fang et al. (2022a, 2022b) improved social-force models to simulate the influence of inclination on passenger walking speed. These methodologies are documented in Appendix H.
4.2. Modeling approaches
Researchers apply various modeling approaches in this research area to formulate the behavior of the problem. The collected papers can be divided into five categories according to the modeling approaches. This paper specifies which methods are more widely employed and offer more significant research advancement opportunities among these categories.