基于IDM与RBFNN的组合型车辆低速跟驰模型.PDFVIP

37 8 Vol. 37 No. 8 第 卷第 期 计算机应用研究 录用定稿 Application Research of Computers Accepted Paper 基于IDM 与RBFNN 的组合型车辆低速跟驰模型* 罗 颖，秦文虎 ( 东南大学 仪器科学与工程学院, 南京 210096) 摘 要：目前针对车辆低速跟驰驾驶的建模研究较少。通过最优加权理论将理论驱动型的跟驰模型与数据驱动型的 (intelligent driver model IDM) (radial basis 跟驰模型进行结合，建立了一种基于智能驾驶者模型 ， 与径向基函数神经网络 function neural network ，RBFNN)的组合型车辆低速跟驰模型。首先对NGSIM 公开数据集进行筛选与处理得到基础 研究数据，之后分别建立基于IDM 与RBFNN 的低速跟驰模型，前者侧重于保证跟驰的安全性与舒适性，后者则能 够输出与真实值更为相符的预测结果，最后通过改进的最优加权目标函数得到最优组合权重，从而建立起了 IDM-RBFNN 组合模型，用平均绝对相对误差MARE 进行了模型的评估，并通过对比分析证明了组合模型具有比单 一模型更优的预测效果。 关键词：车辆低速跟驰；NGSIM ；智能驾驶者模型；径向基函数神经网络；最优加权法 中图分类号：TP391.9 doi: 10.19734/j.issn.1001-3695.2019.03.0052 Combination low-speed car-following model based on idm and RBFNN Luo Ying, Qin Wenhu (Southeast University, School of Instrument Science Engineering , Nanjing 210096, China) Abstract: So far, there are only a few researches focused on the low-speed car-following model. This paper proposed a combined low-speed car-following model based on Intelligent Driver Model (IDM) and Radial Basis Function Neural Network (RBFNN) . This model combines the theory-driven car-following model and the data-driven car-following model by optimal weighting. The first step of the research is processing the public dataset named NGSIM in order to obtain the low-speed car-following data, and then respectively establishing the low-speed car-following model based on IDM and RBFNN. The former focuses on ensu