Research On Dynamic Route Planning Model And Method For Crowd Evacuation

In recent years people's material and cultural life has been greatly enriched along with the rapid development of the economy,and then the assembly in the large public space is also growing.The high density and unstable human flow in the comparatively closed and confined space is easy to lead to some public security incidents,so it has attracted more and more attention from some scholars.Due to that pedestrians are unfamiliar with the environment or there are not any effective emergency management measures,the phenomenon of congestion happens from time to time,and the congestion will further result in some trampling accident.It is significant to carry out some research on the information affecting the evacuation efficiency and optimize the strategy for emergency evacuations to effectively prevent and reduce losses in emergencies.In general,optimizing the emergency evacuation strategy to guide the crowd's evacuating is essential to improving the efficiency of an evacuation,and the better guidance strategy can make the better use of environmental resources(streets,exit,etc.)to shorten evacuation time.One of the effective evacuation guidance strategies is to dynamically plan the evacuation path for pedestrians depending on the real-time evacuation information.Planning better evacuation routes for pedestrians to guide crowds' moving can fully utilize all environmental resources,and contributes to decompose the regional congestion and reduce the occurrence of trampling events.According to the above analysis,we have first analyzed the environmental information affecting planning pedestrian paths in this paper and further studied the dynamic planning path method for evacuations.It has finally proposed a dynamic planning method based on the information sharing algorithms to plan an evacuation path for each pedestrian to guide their moving.The main work and innovations are as follows:(1)It has proposed an information dissemination method to build an information acquisition model,and then analyzed the influence of the incompleteness information caused by the loss of information in its dissemination on planning evacuation routes for pedestrians.The method is built on the “mutual help phenomenon” among pedestrians,and this phenomenon is obtained by analyzing some existing evacuation videos.In this method,it has modified the classical social force model to introduce the attract force from neighboring pedestrians and the driving force from disseminated exits' information to co-drive pedestrians' moving.Because the method simulates that a pedestrian obtains evacuation information directly by itself or indirectly from neighboring pedestrians to choose a proper route and the indirect information is incomplete,it is used to analyze the impact of the incomplete information on planning pedestrians' evacuation route.(2)It has presented a method for evaluating paths to construct a model to acquire complete evacuation information and then utilized them to evaluate all available evacuation routes for crowds to analyze the impact of the complete information on planning evacuation routes.In this method,the traditional social force model is modified to obtain some continuous evacuation routes which represent some evacuation experience from crowds under emergencies.All continuous evacuation routes are discretized,and the discretized paths are optimized by a path learning method to build a candidate path set.The method evaluates all paths in paths set according to the distance from pedestrian to a pedestrian,the length of their path,their congestion in real-time,and the evacuation ability of all exits,and then a pedestrian chooses a candidate path from the path set by evaluating all routes.The influence of all evacuation information on the planning path is analyzed by adjusting their weight parameter of all information which is used to evaluate a candidate path.(3)It has proposed a method to evaluate evacuation efficiency through a series of intermediate evacuation state,and then builds a model to assess an intermediate evacuation state to analyze the influence of intermediate evacuation state on planning evacuation path by reinforcement learning.In this method,the distribution of pedestrians who select the different exits and the equilibrium degree of distribution of all pedestrians is respectively taken as the intermediate evacuation state and its evaluation standard.Base on that,the method further discusses the relationship between the intermediate state and evacuation time by establishing a mapping relationship between crowd evacuation and reinforcement learning.We finally utilize gradient strategy-based reinforcement learning to analyze the influence of intermediate evacuation state on planning pedestrians' evacuation paths.(4)Through the above studies,a dynamic planning path method based on sharing evacuation information is proposed,and then a multi-agent crowd evacuation guidance framework is designed to guide all pedestrians' moving during evacuation.In this method,all agents work together and utilize the sharing information Q learning algorithm to dynamically plan the evacuation path for pedestrians by sharing the real-time local evacuation information.The sharing information in this study includes the horizontal information(the length and the congestion of all segments,exits' congestion,etc.)representing the scene information and the vertical information(the selection frequency and the estimated transit time of road segments)representing evacuation experience.Furthermore,the method also can accumulate and optimize the selection frequency and the estimated transit time of road segments by conducting many simulating experiments,and then save them as evacuation experience.To verify the above research methods,a software platform is built to dynamically plan evacuation path for all volunteers participating in evacuation drills by sharing the real-time evacuation information.In this platform,a set of software installed on volunteers' phones is used to obtain the pedestrian's position by Bluetooth beacons which are pre-installed in the scene.By utilizing each pedestrian's position,the server software in the platform calculates various information about the dynamically planning path(density of road segment,exits' congestion degree,etc.),and then it plans the evacuation path for every pedestrian.The evacuation paths are pushed to each pedestrian by the wireless network to guide the pedestrian's evacuating.By this platform,we let volunteers conduct evacuation drills according to three evacuation strategies(autonomous evacuation strategy,shortest path evacuation strategy and information sharing evacuation strategy).The evacuating results show that the dynamical planning evacuation path based on sharing the real-time dynamic information can effectively avoid the congestion in the evacuation,balance the distribution of pedestrians in each exit especially in the later stage of evacuation,make full use of environmental resources,and shorten the entire evacuation time.