检测数据驱动的交通仿真参数标定方法研究  

Research on Detection Data Driven Calibration Method of Traffic Simulation Parameters

作　　者：马文欣 李瑞敏[1] Ma Wenxin;Li Ruimin(Tsinghua University,Beijing 100084,China)

机构地区：[1]清华大学,北京100084

出　　处：《系统仿真学报》2021年第12期2808-2819,共12页Journal of System Simulation

基　　金：国家自然科学基金(71871123)。

摘　　要：为进一步提高交通仿真模型的准确性,使仿真模型能够及时响应交通需求和驾驶行为的变化,开发了检测数据驱动的动态仿真参数标定方法。提出了检测数据与仿真平台的动态交互方法,选取了北京市光华路/金桐东路交叉口和包头市友谊大街连续5个交叉口作为案例进行了研究。针对初始参数集合进行了敏感性分析,基于分析结果选择所需标定的参数,在VISSIM仿真平台中对案例进行了建模并标定了其驾驶行为参数。仿真结果表明,相比于静态参数标定方法和简单规则动态标定方法,所提方法误差最小,但是选择参数的过程耗时相对较长。To improve the accuracy of traffic simulation model and timely response of traffic demand and driving behavior changes, dynamic calibration method of simulation parameters based on detection data is proposed. Dynamic interaction between detection data and simulation platform is proposed.Intersection of Guanghua Road/Jintong East Road in Beijing and five consecutive intersections on Youyi Street in Baotou are Selected as study cases. Sensitivity analysis is conducted on initial parameter combinations. Based on the analysis results, parameters to be calibrated are selected. Models of cases are developed in VISSIM and driving behavior parameters are calibrated. Simulation results show that compared to static calibration and dynamic calibration based on simple rules, the proposed method yields the lowest error but longest time consumption for choosing optimistic parameters.

关 键 词：交通仿真模型 检测数据 动态参数标定 驾驶行为 敏感性分析 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]