关键词: 动态数据驱动,微观仿真模型,状态估计,数据同化,粒子滤波 中图法分类号TP311文献标识码A

Abstract: The existing pre-established theoretical models in traffic simulation are hardly able to accurately predict the trend of traffic state development,a traffic simulation framework based on dynamic data-driven application system (DDDAS) paradigm is put forward．Firstly,a state space model is proposed to conduct a priori state estimate on the basis of microscopic simulation model；then real-time data is dynamically incorporated in the process of simulation to adjust and evaluate the state space model；what’s more,given the nonlinear and non-Gaussian characteristic of the application field,particle filter is chosen as the means of implementation of data assimilation models and related algorithms by refining the key step and improving related procedures to make enhancement of the ability of state estimation．Finally,micro-simulation system named Movsim is taken to realize the framework．Experimental results indicate our proposed framework poses a high accuracy of prediction and is less vulnerable to the impact of measurement error and environmental noise,thus,it provides stronger stability and reliability．

Key words: Dynamic data driven,Microscopic simulation model,State estimation,Data assimilation,Particle filter