城市交通的并行微观仿真研究模式识别与智能系统专业论文.docxVIP

AbstractAbst Abstract Abst ract Microscopic Traffic Simulation(MTS)replicates the real traffic situation through describing each vehicle’S state．It provides an effective platform for evaluating the impact of various traffic control，management and planning strategies．Recently,it has been widely accepted that MTS is worthy of further investigations to obtain more systematic and comprehensive knowledge of road network planning and traffic management especially for large scale traffic scenarios．However,the computational complexity increased sharply、析tll the growth of the road network scale and the simulation time becomes unacceptable for MTS in large—scale traffic scenarios． This dissertation aims to study and develop a parallel microscopic traffic simulation in order to provide an effective but also highly efficient solution for MTS in large traffic scenarios．The central contents of this dissertation are as follows： 1)The microscopic traffic models are investigated at first and Least—Square Support Vector Regression Car-following model is proposed．In order to reduce the computational complexity,the maximum entropy theory is introduced to select typical samples from the training data set．This model is then evaluated and analyzed to test the accuracy and stability． 2)The framework of the parallel MTS is designed based on the domain decomposition strategy．The load estimation based K-way recursive traffic network partition algorithm is proposed to achieve better load balance． 31 111e bi—level communication framework of the parallel MTS is proposed． The vehicle information exchange and route re—assignment are designed． The communication cost is analyzed． 4) A parallel microscopic traffic simulation system is implemented based on KD一50-I lligh performance computer．And its performances ale evaluated and analyzed． Keywords：Parallel computation，Microscopic traffic simulation，car-following model， road network partition II 中国科学技术大学学位论文相关声明中国科学技术大学学位论文原创性声明 中国科学技术大学学位论文相关声明 中国科学技