Design And Implementation Of Emergency Evacuation Traffic Simulation System Based On SUMO Platform

Traffic simulation system is one of the important research directions of intelligent transportation system.With the problem of urban traffic congestion and the evacuation of various emergency events,the problem is becoming more and more serious.How to use computer modeling and simulation technology to simulate emergency evacuation behavior and evaluate evacuation plan has become an important research content in the field of computer and transportation.At this stage,the design and implementation of emergency evacuation traffic simulation system is less at home and abroad.Therefore,this thesis presents the design and implementation of an emergency evacuation traffic simulation system from evacuation demand,evacuation simulation to evacuation scheme evaluation.The main contributions of this thesis are:(1)The emergency evacuation demand is generated by combining the two dimensions of travel mode and travel distance.Starting from the emergency evacuation demand,based on the theory of stochastic effect maximization,this thesis selects the total travel time,total travel cost,income and other related factors that affect residents’ evacuation decision as utility variables,constructs the utility function,and establishes the cross nested logit model of emergency evacuation demand.At the same time,according to the obtained heterogeneity parameters,the origin destination(OD)matrix of evacuation travel demand is generated by od2 ttrips algorithm,and it is transformed into the specific input of evacuation simulation experiment.(2)A signal lamp phase control system based on deep reinforcement learning is constructed.This thesis designs and implements the traffic signal phase control system by using the deep reinforcement learning algorithm.Through setting up relevant experiments,it is verified that the system can optimize the traffic capacity of key intersections in the road network in the case of emergency evacuation,and can dynamically simulate the real situation of the traffic capacity of the road network.(3)The design and implementation of self driving vehicle system based on reinforcement learning algorithm of strategy gradient is completed.This thesis designs and implements the driving strategy of autonomous vehicle based on reinforcement learning algorithm of strategy gradient.And through setting different road network environment for training and experiment.The experimental results verify that the self driving vehicle after training can eliminate the "walk stop wave" phenomenon in the road,and can also significantly reduce the average waiting time at the intersection of traffic lights,so as to improve the traffic efficiency at the intersection.At the same time,it can also dynamically simulate the real situation of road network capacity.(4)The evaluation index of emergency evacuation scheme is constructed by using entropy weight method.By adjusting the traffic signal control schemes at different intersections and the generation ratio of autonomous vehicles in the road network,different real road network conditions in the evacuation process can be simulated.Taking the generated emergency evacuation demand as input,and using the constructed emergency evacuation system to simulate on SUMO simulation platform,we can get the indicators such as average road delay and average road speed of the road network.In this thesis,the entropy weight method is used to establish the comprehensive evaluation index of emergency evacuation scheme based on these original index data,and to evaluate the advantages and disadvantages of different evacuation schemes.