Prediction On Traffic Congestion In Urban Subway Network Using Multi-agent Based Simulation

In recent years,with the accelerating process of urbanization,the contradiction between the growing demand for urban transportation and the limited capacity of transportation has become increasingly prominent.This has led to the severe traffic congestion in cities in China.The congestion caused by residents' commuting behavior during peak period is even more common.Traffic congestion will not only bring great inconvenience to residents' travel,but also cause huge social and economic losses,seriously restricting the development of the city.Rationally planning and scheduling traffic by predicting congestion is an effective measure to ease traffic congestion in the city.In addition,urban rail transit is playing a more and more important role in the urban traffic of our country.While it effectively relieves the pressure of road traffic,it also gradually faces the challenge of congestion.Traffic flow estimation on the road network is mainly concerned with "vehicles," and traffic congestion can be reflected by parameters such as the average speed of the vehicles and the queue length on the road.For subway traffic,the congestion is not actually related to the state of congestion of the vehicle,but refers to a state in which there are too many passengers and the subway carriages are too crowded,causing a reducing of passenger comfort or resulting in a delayed journey.Therefore,to study the space-time characteristics of subway traffic flow,it is necessary to pay attention to the subway passengers and their travel process.However,there are few related studies on subway passenger volume estimation and congestion prediction.To solve these urgent problems,this study proposes a prediction method for urban subway congestion based on multi-agent based traffic simulation.First,collect data on the residence and work place of urban residents and organize them into OD matrix data that reflects traffic demand.Secondly,by calling the Baidu map's geocoding and route transportation services,we can get residents' detailed travel plans.Then,a multi-agent based traffic simulation framework was designed and implemented,and the residents'travel plans were taken as input to simulate the entire process of residents traveling by subway during peak hours.By optimizing the departure time of commuters and carrying out multiple iterations,the traffic system can reach a stable and relatively superior state.Finally,the simulation results under the stable state are output and traffic flow characteristics are extracted therefrom to identify and predict the metro network sections where congestion may occur.The results of the study verify the validity of the proposed method.At the same time,it confirms that the home-work place data can be used for urban transportation research.This study is of great significance to the optimization of individual traveling process and the effective traffic planning and management for the cities.