Research On Crowd Management And Evacuation Optimization Based On Multi-agent Simulation

As urbanization increasing,urban populations worldwide have grown exponentially.As a result,stampede accidents have become more and more frequent,and the losses they caused have become heavier and heavier.According to incomplete statistics,in China over 80 stampede accidents have occurred since 2000,causing more than 4,000 died and 8000 wounded.Therefore,improving the management level before emergencies and offering high-efficiency evacuation strategies after emergencies,has become an important research topic in the field of pedestrian flow and got a lot of attention.To safely manage the pedestrian flow and to optimally design pedestrian facilities,we need to understand the macroscopic and microscopic rule of pedestrians.Only by exploring the mechanism of stampedes in-depth,can we effectively prevent and manage the accidents caused by crowds,which is the fundament of this research.However,as pedestrians'behaviors are complex and nonlinear,factors that may affect pedestrians'behavior are dynamic and stochastic,it is almost impossible to analyze them in mathematic ways.Further,due to the unobservability of emergency,researchers can hardly obtain enough data to analyze pedestrians in emergency.As a result,research findings in the field of pedestrian flow are quite rare.In consideration of pedestrians'safety,it is impractical to conduct pedestrian experiments,additionally the controllability and repeatability of those experiments are limited.As an important means of exploring the complex system,computer-based simulation has become the main method in the area of pedestrian research.In this research,we employ a multi-agent system to simulate the pedestrians and use Bayesian-Nash Equilibrium to model their decision-making process.Agent-Based Modeling and Simulation(ABMS)is a method raised to deal with difficulties in the Complex Adaptive System(CAS).ABMS is a class of computational models for simulating the actions and interactions of autonomous agents(individual or collective entities).It's by defining the agents'interaction rule in microcosmic perspective and observing the multi-agent system's phenomena emerged in macrocosmic perspective to research the relation between the individual and the system.Though the individual's action is very simple,the reaction of the system caused by individuals can be rather complex.Game Theory is a study of strategic between rational decision-makers,and has been broadly applied in many fields such as social science,economic system,as well as military science.The main purpose of Game Theory is to search and analyze the equilibrium,which is a balance achieved in a game when none of the players can gain more by changing its strategy alone.In other words,no one would change its strategy when the equilibrium is achieved.This paper,therefore,adopts a multi-agent system and the Game Theory to formulate the simulation model.In a game,each of the players seeks its goal,which is to achieve the maximized utility with limited disposable resource,in a rational way.Utility is used as a measure of pleasure or satisfaction in economics.The theory of utility maximization is usually used as the basic assumption to analyze economic activities.It should be noted that this assumption doesn't ensure every player has got or will get the maximized utility.Instead,it means we can use this assumption to analyze or predict player's activities in the perspective of utility maximization.In the model,the tunnel is divided into cells,with each pedestrian in a cell receiving a utility based on the distance to the exit and the number of pedestrians in the cell.Then,each pedestrian uses the Bayesian-Nash Equilibrium to search for the target cell with the maximum expected utility,takes collision avoidance action before moving into the target cell and then searches for the next target cell until exits the tunnel.Based on the proposed model,we conduct experiments in many different situations and summarize the rule of pedestrians'movement.According to the experimental data,we offer some suggestion for the crowd management before emergencies and for the emergency evacuation after emergencies.The main work in this paper are listed as below:(1)Research of establishing a multi-agent simulation model of pedestrian flow.Game Theory has been found to be able to properly capture a pedestrian's decision-making processes and interactions,and agent-based modeling methods has been proven to effectively mimic pedestrian movements.This paper has combined the multi-agent system with Game Theory to research the dense pedestrians in public places.Different from current modeling methods--pedestrians are completely homogenous in the dynamic flow models while partially homogenous in the social force models,cellular automaton models and lattice gas models--in the agent-based models,pedestrians are completely heterogeneous,which means agent-based models can fulfill more accurate simulations,and consequently obtain more convincible result.The proposed model sets the cell width and the time step according to pedestrians'average velocity and step length in the real life.In the model,one or more pedestrians are allowed to share one cell,but the utility each pedestrian gets decreased with the number of pedestrians in the cell increasing.Bayesian-Nash Equilibrium is employed to analyze the pedestrians'decision-making process in depth.The interaction rule between agents and the collision avoidance strategy are well dealt in the model.After finishing the model,it has been calibrated by the Lane Formation and comparing the experimental data to the real data.Based on the calibrated model,this paper summarizes the regulation of pedestrians'movement after conducting simulation experiment in different scenarios,and offers theoretical foundation for the managing of crowds.(2)Research of the crowd management.One of the main purpose of this research is to determine a scientific,practical method for managing crowds at a safe level to prevent possible emergencies.Many researchers have pointed out that emergencies can easily occur when pedestrian density is greater than 4ped/m~2.To be more specific,in a given place,the ratio that the total number of pedestrians to the acreage of the place should be better not exceed 4ped/m~2.However,it lacks feasibility using this static pedestrian density indicator to manage the dynamic crowds,for pedestrian flows are always dynamic and uneven in the real life.Which means,even in a safe pedestrian flow,the density of pedestrian can be greater than 4ped/m~2 in some parts and in an emergency,the density of pedestrians may under 4ped/m~2.Therefore,this paper regards the density of 4ped/m~2 to be the Critical Emergency Density and defines the maximal pedestrian flow rate before the critical density as the Safe Pedestrian Flow Rate(SPFR).In other words,if timely measures are not taken when the pedestrian flow exceeds the SPFR,it is more likely that an emergency could occur.After literatures investigating,fields researching and experiments running,this paper turns the static indicator into a dynamic one,which makes the research result more practical.(3)Multi-scenario research of the pedestrian evacuation under emergencies.Current research seldom combines emergencies with the evacuation.That is to say,specialized research in different situations are needed.This paper classifies the emergencies into static ones and dynamic ones based on their characteristic and discusses the transmission speed and pattern of emergency information.Then makes the rule of pedestrian responding to the emergency information according to key factors in the emergency evacuation.After that,multi-scenario simulation experiments are conducted under static emergencies and dynamic emergencies respectively.Suggestions are then given for evacuations in different situations according to the experimental data,which are beneficial for the safe evacuation in public places.The main conclusions of this paper are as follows:(1)We find expanding the width of pedestrian walkway by one meter will allow the safe pedestrian flow rate to increase by about 3 pedestrians per second.In other words,the ratio that the safe pedestrian flow rate to the tunnel width should better not exceed 3 in order to decrease the risk of emergence.As the pedestrian walkway is a fundamental part of public facilities,this conclusion can be used widely.For crowd managers,they can estimate the safe pedestrian flow rate by measuring the walkway's width.For public place designers,they can optimally design the walkway's width with the predicted pedestrian flow rate.Since the emergency management for crowds is a social science,which is mainly based on experience,our work can give references for the mangers and be beneficial for the public safety.(2)Before emergency occurring,pedestrian managers should pay attention to the status of pedestrian flow's movement,especially in the center part of the walkway,in order to earn more available safe evacuation time for pedestrians.After emergency occurring,for small-scale static emergencies,pedestrian managers should guide all the pedestrian keep walking in their original directions;while for large-scale emergencies and dynamic emergencies,pedestrian managers should guild all the pedestrians moving back before achieving the dangerous area.(3)The following behavior and herd behavior in pedestrian flows have little positive influence during emergency evacuation.From this perspective,crowd managers should take actions such as increasing the spread speed and broadening the influence area of the emergency information,to avoid the following behavior and herd behavior occurring in pedestrian flow as much as possible,and therefore improvig the efficiency of evacuation.The novelties of this paper are listed below:(1)Optimization of pedestrians'microscopic decision process and moving rule.In most microscopic models,conflicts occur when pedestrians make decisions or move at the same time,as each cell can contain only one pedestrian.If conflicts are not dealt with properly in the model,pedestrians go directly through other pedestrians or some other uncontrollable phenomena occurs,which inevitably leads to a false conclusion.Besides,in most research,pedestrians in models are assumed to make decisions one by one,which means pedestrians can investigate others'decisions before deciding and moving.Consequently,these research adopt Nash Equilibrium,which is achieved in the game of complete information,to analyze pedestrians'decision-making.Obviously,in most occasions,this assumption doesn't keep in touch with reality.As current simulation model cannot mimic pedestrians in high densities and collisions between pedestrians are hard to deal with,this paper,therefore,adopts a multi-agent system and the Bayesian-Nash Equilibrium to formulate the simulation model.The Bayesian-Nash Equilibrium is achieved in incomplete information games in which each player makes its decision individually and simultaneously.As in the model,each of the pedestrians has no idea what decisions others may make and as a consequence chooses a strategy which would gain as much utility as possible according to its belief.That is,before moving,each pedestrian has to calculate the expected utility of every possible decision based on the probability distribution of the other pedestrians'choices,after which they choose the cell with the maximum utility as the target cell.(2)Introducing the idea of early warning to the crowd management.After investigating recent pedestrian research,this paper classifies them into pre-emergency research and post-emergency research.Pre-emergency research focuses on the work before emergencies,such as crowd management,risk analysis,public space design and so on,while post-emergency research keeps an eye on the work during or after emergencies,such as evacuation planning and optimizing,loss prevention and so on.It can be summarized from recent research that most of them are focusing on post-emergency evacuation reproduction or optimization,with very few examining situations before the crises occurred.Moreover,crowd management to avoid accidents has received very little research attention,which is of equal importance.This paper comes up with the model that combining multi-agent system with the Bayesian-Nash Equilibrium.In the model,a pedestrian is assumed to choose the target cell from six possible cells and that each of the six alternative cells has eight neighboring cells,which means that many pedestrians may choose the same target cell.In each time step,a pedestrian calculates the utility of all available choices based on their current position and surrounding information.After that,they choose a cell with the maximal expected utility as the target cell.Then they make a judgment as to whether a collision is possible and take action to avoid the collision.Finally,they move into the chosen cell and begin the next decision-making round until they pass through the tunnel.It's by simulating pedestrians microscopic this paper investigates the phenomena emerged in the macroscopic.Based on the experimental data,the paper summarizes the regulation of their movement and offers suggestions for crowd management in order to reduce potential losses.(3)Optimization research of multi-scenario evacuation strategy.This paper classifies emergencies into static ones and dynamic ones according to their attributions.Dynamic emergencies are emergencies with influenced area broadening over time while static ones are not.In the improved model,the information transmission path and the pedestrian responding ratio are fully considered.Then,simulation experiments are conducted under different situations.Based on the experimental data,we have concluded that for small-scale static emergencies,pedestrian managers should guide all the pedestrian keep walking in their original directions;while for large-scale emergencies and dynamic emergencies,pedestrian managers should guild all the pedestrians moving back before achieving the dangerous area.This conclusion has provided a scientific reference for emergency management and can efficiently shorten the required safe evacuation time,which will beneficial for the evacuation optimization by reducing losses as much as possible.