基于卡口系统的路段旅行时间预测及系统原型设计

发布时间：2018-03-11 21:28

本文选题：卡口系统　切入点：旅行时间估计　出处：《山东大学》2015年硕士论文　论文类型：学位论文

【摘要】：近年来,城市道路交通压力不断膨胀、网路交通需求急剧增长,使得城市道路拥堵也日益加剧,时有伴随局部交通瘫痪等问题。交通拥堵不仅给出行者造成时间延误和经济损失,还给整个社会造成了潜在的交通风险、能源损失、以及环境污染,严重影响了城市和城市交通的发展。智能交通系统的出现为缓解或解决这一城市痼疾提供了新方法和思路。而作为其重要构成部分之一,城市路段旅行时间预测,是交通规划、交通运营与管理、交通态势监控、交通诱导、交通事件处置和出行信息服务等一系列举措的核心基础问题之一。该论文通过查阅国内外相关文献和研究成果,总结现有方法的优缺点,从交通系统的特点出发,对城市路段旅行时间预测方法展开研究,为交通管理者和驾驶员提供数据支撑和决策依据。首先,分析了当前交通信息采集技术原理及优缺点,研究了使用卡口系统数据的原因,介绍了其采集的数据特点。以此数据为基础,分析了错误数据和缺失数据出现的原因,提出了相应的处理和修复方法,进而得出了旅行时间估计值。然后,以济南市经十路采集的真实数据为例,研究了旅行时间工作日和非工作日同时段具有较大差异、全天具有显著早晚高峰、以及工作日同时段具有相似性及波动性等特征,基于该类特性,对原有模糊神经网络进行了改进,构筑了稳定性和容错性更强的补偿模糊神经网络预测模型。利用济南市卡口系统实际采集的数据,对比分析了各种预测模型,结果证明：在1.7km的城市路段中,补偿模糊神经预测模型平均预测误差均小于0.4min,相对预测误差小于30%的概率均超过了97%,在准确性和可靠性方面,均优于BP神经网络和卡尔曼滤波算法。最后,设计了旅行时间预测系统原型。整个原型系统包括数据采集与存储层、数据处理层、功能管理层和用户交互层,并给出了实现流程和各层功能模块。
[Abstract]:In recent years, urban road traffic pressure has been expanding and the demand for network traffic has increased sharply, which has made urban roads congested increasingly. Traffic congestion not only causes time delay and economic loss for travelers, but also creates potential traffic risks, energy losses and environmental pollution for the whole society. The emergence of Intelligent Transportation system (its) provides new methods and ideas for alleviating or solving this chronic disease of the city. As one of its important components, the travel time prediction of urban sections is traffic planning. Traffic operation and management, traffic situation monitoring, traffic guidance, traffic incident disposal and travel information service are among the core basic issues. This paper summarizes the advantages and disadvantages of the existing methods, starting from the characteristics of the traffic system, carries out a study on the travel time prediction method of urban road sections, which provides data support and decision basis for traffic managers and drivers. First of all, This paper analyzes the principle, advantages and disadvantages of the current traffic information collection technology, studies the reasons for using the data of the bayonet system, and introduces the characteristics of the data collected. Based on the data, the reasons for the occurrence of the wrong data and the missing data are analyzed. The corresponding treatment and repair methods are put forward, and then the estimated travel time is obtained. Then, taking the real data collected by the tenth way of Jinan as an example, the paper studies the great difference between the travel time and the non-working day. The whole day is characterized by significant morning and evening peak, similarity and volatility at the same time during the working day. Based on this characteristic, the original fuzzy neural network is improved. A predictive model of compensatory fuzzy neural network with stronger stability and fault tolerance is constructed. By using the actual data collected by the Jiaokou system in Jinan, various prediction models are compared and analyzed. The results show that: in a 1.7 km section of a city, The average prediction error of compensatory fuzzy neural prediction model is less than 0.4 min, and the probability of relative prediction error less than 30% is more than 97. In terms of accuracy and reliability, it is better than BP neural network and Kalman filter algorithm. The prototype of the travel time prediction system is designed, which includes data acquisition and storage layer, data processing layer, functional management layer and user interaction layer.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491;TP274

【相似文献】