Vulnerability Research And Recovery Strategy Of Emergency Evacuation Paths In Sports Venues

In recent years,due to the rapid development of the economy and the continuous improvement of China’s internationalization level,more and more sports events and large-scale social public activities are held in various cities,and their scale is also constantly expanding.The types of activities undertaken by large sports venues are also gradually increasing,which has higher requirements for the number of spectators they can accommodate and also poses greater challenges to their safety performance.In this situation,the emergency evacuation network of large sports venues bears a relatively important responsibility,and the harm and impact of unexpected events are increasingly valued.In order to enable large-scale sports venues to evacuate spectators in a timely manner and reduce the occurrence of accidents in emergency situations,it is particularly important to study their emergency evacuation routes.This study is of great help in improving the safety performance of sports venues and is of great significance for maintaining the normal operation of sports venues.The paper is based on the topology rules of complex network theory,based on the relationships between nodes,and applies the Space L method to abstract the evacuation path of large sports venues into a network topology diagram.The average path length,clustering coefficient,and traffic contribution rate of the road network are calculated.Based on the above analysis,two attack modes are implemented on the network,namely random attack and deliberate attack,identifying key nodes,and analyzing the vulnerability of the road network.Secondly,based on graph neural networks and reinforcement learning theory,encoding is constructed using the Graph SAGE algorithm,where nodes and network states are expressed as low order feature vectors.The decoding process is constructed using the Q-Learning algorithm,and a road network recovery strategy algorithm framework based on "encoding decoding" is established.By analyzing examples and comparing the three recovery strategies composed of node degree recovery strategy,road network traffic contribution recovery strategy,and graph network combined with reinforcement learning,it is concluded that graph network and reinforcement learning strategy can restore the traffic and smoothness of the network in a shorter time to meet evacuation requirements,and graph network and reinforcement learning strategy are the optimal recovery strategies,And analyze the selection order of recovery sequence nodes considering both recovery costs and not considering recovery costs.In this paper,the concepts of road network traffic degree and node traffic contribution rate are proposed to effectively evaluate the vulnerability of road network.A road network recovery strategy algorithm based on the "encoding decoding" structure was constructed,and network information was deeply mined using graph neural network methods.At the same time,considering the cost of node recovery,combined with reinforcement learning methods and case analysis,it was found that for damaged network structures,this method can achieve good network recovery results.