Indoor Crowd Evacuation Modeling And Decesion Optimization Based On Group Chase And Escape

Group chase and escape is a widespread group game phenomenon,which means that two or more different types of individuals are engaged in chase and escape actions,and make up the phenomenon of collective flow movement.The group chase and escape movement originated from real natural scenes.In the early days,it attracted the attention of a large number of scholars because of the widespread animal hunting behavior in nature.At present,the research of group chase and escape has penetrated into many fields such as terrorist attack drills and crowd evacuation.In the scenario of group chase and escape,the trajectory and dynamics between groups are closely related to the strategies adopted by the participants.Exploring different strategic choices of the participants in the group chase group chase and escape,and optimizing its intelligent decision-making process,has universal reference value in many scenarios,such as the intelligent evacuation process of crowds.In traditional crowd evacuation scenes,there is generally only one type of role(evacuation pedestrian).And usually the simulation of such scenes only needs to consider the path finding problem and the collision avoidance problem.In the model design of this paper,mobile attackers who pose a huge threat to pedestrians during the evacuation process are also taken into account,that is,a crowd evacuation environment under the condition of group chase and escape is formed.Although the probability of terrorist attacks is very small,modeling the evacuation environment with mobile attackers still has a certain preventive effect and research value.This paper will mainly model and simulate the crowd evacuation in the crowd chase scene,and set the evacuation site in the complex indoor environment on the first floor of Joy City in XiDan,and establish a continuous space evacuation model based on this.In the model designed in this paper,the crowd in the group chase and escape scene needs to complete two tasks:avoiding the attacker and evacuation to the safe exit.Sometimes the two can be completed at the same time,and sometimes there will be a conflict,which means that pedestrian needs to make a compromise choice between the two tasks at this time.Based on this assumption,this paper studies the overall time-consuming and casualties of the entire evacuation process when pedestrians take into account both rapid evacuation and avoidance of danger.Through many experiments,this paper found that pedestrian decision-making in the group chase scene is very important to the evacuation result,and the simple artificial strategy often fails to achieve a good evacuation effect.Therefore,this paper combines some deep reinforcement learning algorithms,uses the models such as Deep Deterministic Policy Gradient(DDPG)network and Bidirectionally-Coordinated Nets(BicNet)to optimize the decision-making process of pedestrians to allow evacuation pedestrians make intelligent decisions based on specific scenarios that appear during the evacuation process,which improves the efficiency of indoor evacuation and reduces the overall evacuation time and the number of casualties.