面向多源数据融合的高速公路检测器布设方法研究

发布时间：2018-06-16 11:02

本文选题：手机切换定位 + 行程时间估计　；参考：《东南大学》2015年硕士论文

【摘要】：交通检测器的布设决定着路段行程时间估计及预测的精度,行程时间估计方法的改进也反作用于检测器布设方法的研究与应用。行程时间作为高速公路交通管理的关键参数,可以为交通诱导方案的实施和出行信息的发布提供数据支持和决策参考。根据对现有研究文献的总结,面向数据融合研究高速公路检测器布设方法,将手机定位交通信息提取作为检测器布设考虑的因素作为本文交通检测器布设研究的一个创新性出发点和切入点,研究基于手机切换点提供交通状态数据的支持下,利用BP神经网络数据融合方法估计行程时间,并以行程时间估计误差为约束反推研究固定检测器布设方法。首先,以手机定位交通信息提取作为研究的切入点和创新点,介绍手机定位技术在交通信息提取方面的应用,着重分析基于手机切换定位技术的交通信息提取系统框架和技术流程。基于手机切换定位技术特征,与传统固定检测方式和传统GPS浮动车检测技术进行比对分析,验证手机切换交通信息提取的技术优势和应用前景。同时,参考传统GPS浮动车微观仿真常用方法,结合VISSIM仿真固有特点,提出手机切换定位交通流微观仿真方法,通过对实际交通情况仿真和结果分析加以验证。研究表明,在手机切换样本比例足够(不低于3%)条件下,基于手机切换数据的行程时间估计精度比较准确,且不易受交通运行条件的影响。通过对实际路段交通场景的仿真获得固定检测数据和手机切换数据,采用直接估计方法和融合方法估计路段行程时间。研究提出采用BP神经网络方法融合固定检测数据和手机切换数据用以估计路段行程时间,以满足交通状态不稳定的估计需求,并建立行程时间估计精度评价指标。根据研究需要和拟定的数据融合流程,分别对全路段和切换路段进行融合估计,结合误差指标对比分析单一数据源和多数据源进行行程时间估计,并进一步对融合结果完成深入分析。研究表明,基于BP神经网络的行程时间融合估计效果相较单一数据源,可以得到更精确、稳定的结果和效益,且交通条件差时,优势更明显。结合检测器布设所需要考虑的原则与影响因素,在传统检测器布设经验思路的基础上,以手机切换交通信息采集为支撑环境,采用BP神经网络融合方法进行基于多源数据的行程时问估计,以误差和精度作为检测器布设的约束与目标。确定了均匀布设方案和盲区补充方案两种检测器布设策略,结合路段实际交通环境进行微观仿真。根据估计指标确定方案的可行度,并针对各可行方案的误差指标对比分析。对实例进行验证,得到最优方案。研究表明,检测器布设间距在1000-1500m附近时,行程时间估计误差变化趋平,布设方案较为合理；且应考虑在切换长度较长、有出入匝道处等“盲区”布设检测器,以实现成本受限情况下的估计效益最大化(即误差最小)。
[Abstract]:The layout of the traffic detector determines the precision of the travel time estimation and prediction, and the improvement of the travel time estimation method is counterproductive to the research and application of the detector layout method. As a key parameter of expressway traffic management, travel time can provide data support and decision reference for the implementation of traffic guidance scheme and the release of travel information. According to the summary of the existing research literature, the method of highway detector layout for data fusion is studied. Taking mobile phone location traffic information extraction as the factor of detector layout as an innovative starting point and a breakthrough point of this paper, the research based on mobile phone switching point to provide traffic state data support, The BP neural network data fusion method is used to estimate the travel time, and the fixed detector placement method is studied with the travel time estimation error as the constraint. First of all, take mobile phone location traffic information extraction as the breakthrough point and innovation point, introduce the application of mobile phone location technology in traffic information extraction, The framework and technical flow of traffic information extraction system based on mobile phone switching and location technology are analyzed. Based on the characteristics of mobile phone switching and location technology, compared with the traditional fixed detection method and the traditional GPS floating vehicle detection technology, this paper verifies the technical advantages and application prospects of mobile phone switching traffic information extraction. At the same time, referring to the traditional micro simulation methods of GPS floating vehicle, combined with the inherent characteristics of VISSIM simulation, the microscopic simulation method of mobile phone switching and positioning traffic flow is put forward, which is verified by the simulation of the actual traffic situation and the analysis of the results. The results show that under the condition that the sample ratio of mobile phone switching is enough (no less than 3), the estimation accuracy of travel time based on mobile phone switching data is accurate, and it is not easy to be affected by traffic operation conditions. The fixed detection data and the mobile phone switching data are obtained by simulation of the actual road traffic scene. The direct estimation method and fusion method are used to estimate the travel time of the road section. In this paper, BP neural network method is proposed to combine fixed detection data and mobile phone switching data to estimate the travel time of road sections to meet the needs of unstable traffic conditions, and to establish an evaluation index of travel time estimation accuracy. According to the research needs and the proposed data fusion process, the fusion estimation of the whole road section and the switching section is carried out, and the travel time estimation of the single data source and the multiple data sources are compared and analyzed in combination with the error index. The fusion results are further analyzed. The results show that the estimation effect of BP neural network based on travel time fusion is more accurate, stable and effective than that of single data source, and the advantage is more obvious when traffic conditions are poor. Combined with the principle and influence factors of detector layout, based on the experience of traditional detector layout, the mobile phone switching traffic information collection is used as the supporting environment. The BP neural network fusion method is used to estimate the travel time based on multi-source data. The error and precision are used as the constraints and targets of the detector placement. The uniform layout scheme and the blind area supplementary scheme are determined, and the microscopic simulation is carried out in combination with the actual traffic environment. The feasibility degree of the scheme is determined according to the estimation index, and the error index of each feasible scheme is compared and analyzed. The example is verified and the optimal scheme is obtained. The results show that when the distance between the detectors is about 1000-1500m, the error of travel time estimation tends to flatten, and the layout scheme is reasonable, and it should be considered that the detector should be arranged in the "blind area" such as the long switching length and the entrance and exit ramp, etc. In order to maximize the estimated benefit (I. e., minimum error) under the condition of limited cost.
【学位授予单位】：东南大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491.116

【相似文献】