Research On Crowd Evacuation Behaviors Based On Evolutionary Game Theory

How behaviors evolve is considered as one of the top twenty-fivescientific problems by the well-known journal Science. Driven by internalself-organized behaviors and external centralized control behaviors, thecrowd's evacuation behaviors exhibit complexity of evacuation process.In this work dynamic floor field was firstly studied based on cellularautomata model, and it was extended to establish an information fieldmodel, and then the information field model was used to study the controlproblems on crowds' behaviors in terms of evacuation guides. Andsecondly an escape game model was proposed to discuss the impact ofescape aspiration on the cooperation in the evacuation process with therules of behavioral evolution determined by reinforcement learning.Lastly, by two-strategy escape game and three-strategy conflict gamerespectively, rationality degree and herding degree were primarily studiedon the evolution of evacuation behaviors and the evacuation efficiencycompared with the latest psychological research findings.In this work the evolutionary game theory is introduced to researchon crowd evacuation, and the mechanism of cooperation in the evacuation crowd has been found as follows.(1) In an emergency evacuation thefrequency of cooperation will increase with the increasing degree ofimitation among evacuees as well as the escape aspiration.(2) Duringemergency evacuations, the people with lower-rationality prefer tocooperate, and the evacuees with hyper-rationality are likely to vie withothers and spend more evacuation time.(3) The conflicts between theevacuation crowds at bottleneck indicate a process of game, which meanscooperation during emergency evacuation is not derived from individualmoral standard but the self-regarding behaviors. The above interestingfindings provide us with a new insight into crowd's evacuation behaviors,which also supports the observation from the psychological research.Additionally, in this work another research is focused on the lack ofclassic cellular automata and a new heterogeneous algorithm to computedynamic floor field is proposed by revising the original model. Aninformation field model with different mechanism of transmission ofinformation is established inspired by the mechanism of spread of thedynamic floor field. Under the framework based on the information fieldmodel, effective and precise information transmission by setting propernumber of guides can improve evacuation efficiency significantly.Therefore, the information field model which can be integrated with theoriginal floor field model, as a new research method, is able to provide usa new understanding on the evacuation environmental conditions The innovation of this work is represented as the followings.(1) Themethods such as evolutionary game theory and game learning theory arefirst applied to the evolution of behaviors of large-scale evacuation crowdwith the simulation results in good agreement with the latestpsychological research findings. This quantitative approach is likely to begenerally used to further study, which can transform the notions of crowdbehaviors scientifically.(2) The inappropriate computational method ofdynamic field has been modified and an information field model has beenestablished, with which a method for crowd control has been proposed.In this work cellular automata model was adopted to deal with thecomplex evacuation system. More individual factors and complexenvironment factors can be taken into account for the future work andutilizing virtual reality technology to capture the response of behaviors ofparticipants can identify the parameters in the model and provide a moreapplicable framework of evacuation model.