Study On Crowd Evacuation Model And Simulation In Subway Station Under Sudden Fire

With the vigorous development of urban rail transit,the characteristics of convenience,safety and punctuality,subway makes people first choose it as a way of travel,and it is also an important way to alleviate urban traffic congestion.However,the subway station has some characteristics such as irregular layout,closed space,dense passenger flow and fast flow speed.In case of emergency,especially serious fire,passengers often show panic and disorderly behavior,which brings great challenges to the emergency evacuation work in case of station emergencies.If passengers cannot be evacuated orderly and efficiently,and there is no reasonable evacuation guidance,it is prone to serious crowd congestion,even crowd stampede and other dangerous accidents.Therefore,in order to reduce personnel casualty casualties and property losses,it is of great significance to study the moving characteristics of people in subway stations and the planning methods of evacuation route for guides under sudden fires.In this paper,the multi-grid model and reinforcement learning algorithm are taken as the basis to study the motion characteristics of fire personnel,and the path planning method of the guide is designed which is applied to a case.First,this paper analyzes the influencing factors of fire on personnel.It is proposed that fire repulsion is composed of heat generated by fire temperature and visual power caused by smoke.Then an emergency evacuation model is constructed taking into account fire repulsion,by which update rules of the model is defined.Through experimental comparison,the results show that the evacuation model considering fire repulsion can affect the personnel evacuation path and moving speed,and the improved model is more in line with the actual characteristics of crowd evacuation movement.Secondly,aiming at the problem of evacuation route congestion when the guide guides the crowd evacuation in a dynamic environment,the traditional reinforcement learning path planning algorithm is improved.The achieved local optimal point is marked through adding a tabu table to the algorithm,and a global optimal solution is obtained through strengthening the locally optimal solution,the process of frequent trial and error for reinforcement learning is reduced.Then,simulation experiments are carried out in different evacuation scenarios to verify that the path searched by the improved algorithm is the best evacuation path.Finally,the improved algorithm is combined with crowd evacuation simulation based on the case of Dalian subway station,and the environmental model for subway station is established.The simulated evacuation behavior is guided by the movement probability of people under fire repulsion and the guide path planning based on Tabu Search reinforcement learning algorithm.Experiments are carried out under different exit number and width scenarios to further verify that the proposed evacuation model can improve the evacuation efficiency and has certain practical significance.