Proposal of social force model including group behaviour for pedestrian evacuation in emergency situations

Security is one of the basic needs of people. On the other hand, situations such as fire, terrorist attacks and natural disasters pose a major threat to the lives and property of people and other living things. Although we have not met with these kinds of disasters very often throughout our lives, we may encounter very dramatic results depending on whether we are caught unprepared or not. Because these disasters are often impossible to avoid, Disaster countermeasures will be useful in terms of minimizing the losses in case we meet with a disaster. The modeling of pedestrian behavior in emergency situations is important for many disciplines. The modeling of pedestrian behavior is an important factor for the design of new buildings, robot programming, pedestrian traffic planning and emergency planners. Traffic and civil engineers can use these models as an input to design structures that will provide the most efficient evacuation and pedestrian flow. Particularly in countries where the number of pedestrians is increasing, such as Istanbul, the modeling of pedestrian behavior has a critical prescription. Effective evacuation plans are needed to reduce the most casualties in emergencies. Experimental studies and computer based analysis studies related to evacuation planning of structures (buildings) at these points are seen as suitable methods. Evacuation planning involves iterative processes and aims to ensure that people in a particular area are removed from this area as soon as possible in a risky situation and identify the most appropriate routes to provide these exits. Computer-aided analysis studies seem to be used more widely in the literature in evacuation planning, as they are less costly than experimental (practical) studies. Mathematical equations, some physics theories and simulation models are the basis of the computer assisted analysis studies used in the evacuation planning. Models used to simulate pedestrian dynamics have begun to attract much attention in recent years, because these models play an important role in predicting pedestrian behavior and in designing evacuation zones. The models used in modeling the behavior exhibited by the publications are based on physical rules and mathematical equations. According to the results of literature review, there are various approaches in classification of models in studies. Microscopic-macroscopic, homogeneous-heterogeneous, intermittent-continuous and stochastic-deterministic models are the most common types of classifications. The majority of names used in these classifications refer to a state or a feature in a model. In this thesis, models will be introduced and examined under macroscopic and microscopic and mesoscopic titles in classification. When we examine the models collected under the three basic headings, it is seen that there are differences in the complexity of the calculation and the level of detail between the models. While microscopic models require the most computation time, they provide a more detailed way to create realistic pedestrian patterns. On the other hand, where the number of pedestrians is too great, the macroscopic models may become more suitable due to the exponential increases during the calculation. Where interactions between pedestrians are not important, macroscopic models may be a suitable modeling approach. it is seen that the number of simulation software based on the Agent Based modeling approach is higher than the number of other model types. The most important reason for this is that the Agent Based modeling approach can provide more realistic results in modeling human behaviors. Agent-based models define pedestrians as autonomous agents that can interact with each other and their environment. Based on agent-based modeling and simulation, the work is based on complex adaptive systems. Complex adaptive systems have the characteristics of autonomous individuals in the system, interaction between these individuals and adaptation of individuals to changing environment conditions. In complex systems, behaviors that are difficult to predict, which come into play due to the interaction of each individual in the system and their interactions, are called emergent behaviours. Agents are units that have the ability to make decisions in a complex adaptive system. Agents are autonomous individuals with the ability to receive information, process information, adapt to changing environment and make decisions in line with the rules they have. Agent-based modeling and simulation utilizes discrete event simulation and object oriented programming methods to manage complex adaptive systems. It is also referred to as the down-up modeling approach because it allows the generic behavior of the system to be revealed by one-to-one modeling of each agent in the system. The Social force model is based on the assumption that the movements of the pedestrians are facilitated under the influence of certain forces. It is basically based on Newton's second fundamental law of motion. The model received this name because the internal motivations of the pedestrians are considered as forces. Although the social force model is lacking, it is a very valuable model for modeling pedestrian movements. Models in which the movements of the pedestrians in an emergency are modeled as robots, only physical ones as particles, will always be missing. The influence of factors such as psychological, sociological, educational, etc. they have under the behaviors exhibited by the victims in an emergency must not be ignored. In addition, pedestrians tend to interact with other pedestrians around as a social entity. It is seen that the effects of these factors and the behavioral situations arising from these factors must be examined when modeling the movements of the pedestrians in the emergency situations. We can divide the behaviors exhibited by the pedestrians into two main groups: group behavior and individual behavior. It is unacceptable to ignore group behavior in modeling pedestrian behavior. The pedestrian has a tendency to exhibit largely group behavior from individual behaviors in emergencies. In this thesis, a mathematical model has been created using AnyLogic to simulate escaping pedestrians behaviour during an emergency situation. The proposed model is based on the basic SKM model of Helbing and some changes and additions have been made in the direction of the group behavior suggested on Helbing's model in order to make the pedestrians behavious more realistic. The model has additional forces implemented both for improving social force model and to implement group behaviour. Agent-based modeling approach has been used as modeling approach and the pedestrians has been modeled as acting agents acting under specific forces. Depending on the size of the determined step, pedestrians are accelerated under the influence of the forces acting on them and they changes their speed and position depending on this acceleration. Each simulated pedestrian is evaluated as an independet agent that move according to its local perception of the dynamic environment and the movements of the pedestrians are obtained by solving the ordinary differential equation which is formed by the forces acting on them by the Euler method. While the generated model is continuous in the two dimensional plane, time is discrete due to the numerical approach used to solve the ordinary differential equation. In this work, the pedestrians are divided into groups and they allow to move in groups. The social distance theory and the K-Means clustering algorithm have been utilized while building pedestrian groups. In cases where the pedestrian density is less than 0.125, the pedestrians are grouped according to the social distance theory and the K-means clustering algorithm is used where pedestrian density is greater. Inspired by the Boids algorithm to calculate the perceived mass centers of the pedestrians in the group, and by applying an imaginary group force to the group members from the center of mass, the group members were made to move close to each other. In addition, the leading attractor force was added to the social force model in order to examine the leader behavior within each group to examine the behavior of the leader, a subtype of group behavior. A leader driving force is also added to prevent the leader from going to the target. Modeling of pedestrians with different size and number of group members is provided. The creation of realistic pedestrian groups has been achieved with the help of the theory of social distance utilized in the formation of groups. In case of emergency, the feeling of hastiness of the pedestrians is included in the model by the desired speed being dependent on the time. Simulation experiments were carried out to demonstrate the effectiveness of the proposed model. Different scenarios such as pedestrian number, exit number, exit width, presence of obstacles, location, group number and leader number were tested and evacuation times were obtained for all scenarios. Obtained evacuation times were obtained by taking Monte Carlo approach and taking averages of at least 100 runs with a margin of error of 0.05 in 95 confidence intervals. To verify the proposed model proposed, two apprhach was applieds. First, the model was compared with the studies in the literature and it was seen that the results gave similar results to the studies. Then, the situation of bringing the pedestrian phenomenon that emerged in an emergency into the model was examined, and it was observed that most of these cases were revealed. In the further studies we aims to create more realistic pedestrian models with new models to be created based on pedestrian data analysis and fuzzy logic. We also consider to compare our model with some real evacuation experiments for validation.