【作者】 杨达；

【导师】 蒲云；

【作者基本信息】 西南交通大学 ， 交通运输规划与管理， 2013， 博士

【摘要】 随着经济的发展和制造业成本的下降,社会车辆保有量目益增加,而道路的供给增加速率远远赶不上车辆的增加速率,造成了严重的道路供需不平衡问题,这个问题在像我国这样的发展中国家,表现更加明显。此外,社会上车辆保有量的增加也加剧了环境的污染,增加了交通事故的发生率。在这种背景下,可以有效缓减交通拥堵、降低交通事故发生率并减少环境污染的新技术——ITS (Intelligent Transportation System)技术得到了很大的关注和快速地发展。在ITS技术中,一项受到世界各国学者重视的新技术——车联网技术(Connected Vehicle Technology, CVT或者是Internet of Vehicles,IOV)正在对交通环境产生巨大影响。在车联网环境下,车辆和车辆之间、车辆和交通设施之间、交通设施和设施之间的通信大大增强,驾驶员能接受到的外界刺激也随之大大增加,会使得驾驶员的驾驶行为发生重大变化。本论文关注的是在车联网环境下,车联网技术对车辆跟驰行为产生的影响,试图为新技术的到来在车辆跟驰行为理论上进行一定的准备。新的交通环境对车辆跟驰行为最直接的影响是在行驶过程中后车的信息得到了很大的加强,驾驶员的行为受到后车的影响较之以前更加明显。所以本文试图建立考虑后车的车辆跟驰模型,并在模型的基础上分析考虑后车的车辆跟驰行为的特点。此外,随着自动控制理论在交通领域的应用,特别是ACC (Adaptive Cruise Control)技术的发展,很多学者在寻找更为合适的车辆行驶控制机理,本文提出的考虑后车的车辆跟驰模型可以为ACC设计提供—种可能的控制机理。在对考虑后车的车辆跟驰模型进行分析时,发现当前车辆跟驰研究领域存在着-个重要的问题：当前车辆跟驰模型的标定方法标定出来的参数在仿真中会引起一些不真实的交通现象,所以本文在建立考虑后车的车辆跟驰模型并对其进行标定之前,首先提出了两种对当前车辆跟驰模型标定进行改进的方法,第一是对不合理交通现象(包括过大的加速度和减速度、超过限速的速度和过高频率的撞车)的惩罚方法,第二是误差权重法,主要是针对在车辆跟驰行为仿真中随着迭代次数的增加仿真误差不断积累的问题。选用了相对速度、安全距离和理想速度三大类车辆跟驰模型中的五个车辆跟驰模型GM, Bando, Gipps, FRESIM和IDM,并使用了著名的微观交通流实际数据——NGSIM (Next Generation Simulation)数据对这两种改进的方法进行评价。本文提出了一个考虑后车的车辆跟驰行为的建模框架,该框架可以将当前大部分的车辆跟驰模型转化为考虑后车的车辆跟驰模型。根据该建模框架,将Chandler, GM, Bando, Jiang, Gipps和NETSIM六个模型从当前形式发展为考虑后车的形式,建立了他们的考虑后车的车辆跟驰模型,并对这六个模型使用NGSIM数据进行了参数标定和在不同的后向控制比例下的加速度特点分析。作为车辆跟驰模型的一个十分重要的特点,本文对车辆跟驰模型的稳定性也分别使用理论分析法和基于真实数据的仿真法(虚拟环形路法)进行了分析,其中对考虑后车的Gipps车辆跟驰模型使用的是理论分析的方法,而对相对速度模型和理想速度模型使用的是基于实际数据的仿真分析法。通过分析,本文得到了如下的几个结论：1.本文所提出的改进的车辆跟驰模型标定方法——对不合理交通现象惩罚方法和对积累误差进行限制的误差权重法,通过仿真实验证明是有效的；2.不同的车辆跟驰模型在标定中可能会有不同的最优目标函数,但是当模型的最优目标函数不明确时,以位置为变量的加权MAE(Mean Absolute Error)在大部分模型中可以获得精度比较高的标定结果,而以速度为变量的加权U Theil’s函数在大部分模型中可以比较有效地控制仿真误差的积累；3.考虑后车的车辆跟驰模型比不考虑后车的车辆跟驰模型有更高的精度,也就是说考虑后车的车辆跟驰模型在一定程度上更能反映真实交通流的情况；4.在当前交通环境下的车辆跟驰行为中,前向控制占据了80%以上,而后向控制所占的比例比较小,而且大部分模型显示后车迫使前车加速的情况更为常见；5.随着后向控制比例的增加一些模型的前向后加速度分布会变得不规则,说明过高的后向控制比例会引起车辆跟驰行为的混乱,所以在对车辆行驶进行控制时,要将后向控制的比例限制在一定的合理范围内；6.本文得到的考虑后车的车辆跟驰模型的线性稳定性条件通式通过证明是正确的,它适用于在决策变量中不包含反应时间,决策变量为加速度,而且表达式相对简单的考虑后车的车辆跟驰模型；7.考虑后车的Gipps车辆跟驰模型显示后向控制对交通流有三种效果：使交通流变得稳定、使交通流变得不稳定、使交通流变得不合理,该结论丰富了当前有关后车对交通流稳定性影响的研究成果；8.基于实际数据的虚拟环形路仿真显示,自由交通流在加了扰动以后在稳定性测试中可能会发生严重的交通拥堵现象,而拥堵交通流在加了扰动以后也有可能达到自由流状态,虽然可能伴随着微小的波动；9.通常来说在自由流中可能没有必要考虑后向车辆对车辆跟驰行为的影响,而且后向控制的比例在总的决策中可能需要被限制在一个合理的范围内来避免交通协调和控制的失效。更多还原

【Abstract】 With the development of economy and the reduction of the manufacturing cost, the number of vehicles increases very fast. However, the increase of roadways cannot catch up the increase of vehicles, that is to say, there is an imbalance of roadway supply and demand. This problem is much more serious in the developing countries like China. In addition, a large number of vehicles also cause a lot of environmental pollution problems and traffic accidents. To solve these problems, a new emerging technology, ITS (Intelligent Transportation System), has attracted much attention and been developing rapidly in the past decades. As a critical technology in ITS, CVT (Connected Vehicle Technology) is gradually changing driving environment in the whole world. In CVT environment, the communications among vehicles, between vehicles and traffic facilities, and among traffic facilities are greatly enhanced, which reveals that drivers can receive much more external stimulation from other vehicles and facilities, and the drivers’ behavior will be influenced hugely.This paper focuses on the influence of the new vehicle connected technology on the car-following behavior and tries to prepare the car-following theory for the coming change of the traffic environment. It is intuitive that drivers can receive much more information from the following vehicle in the new traffic environment, and drivers’behavior will be definitely impacted by this change. Thus, this paper attempts to model the car-following behavior considering the following vehicle and analyzes the characteristics of the car-following behavior considering the following vehicle based on the proposed model. On the other hand, along with the application of the automatic control theory in transportation field, especially the ACC (Adaptive Cruise Control) technology, a lot of scholars have been searching for the more reasonable vehicle control mechanism, and the new proposed car-following model may provide ACC a possible better control mechanism.In study on the car-following model, it is found that the existing car-following model calibration method has some problems, such as producing impractical traffic phenomenon and high prediction error. Therefore, the existing calibration method needs to be improved before applied to calibrate the car-following model considering the following vehicle. Two methods are introduced to improve the existing calibration method in this paper. The first one is the punishment method, aiming to punish the abnormal traffic phenomena (including excessive acceleration and deceleration, exceeding the speed limit, and frequent crashes) in car-following simulation. The second one is the error weighted method, which is mainly to against the problem that the simulation error accumulates with the increase of the simulation iteration number. These two methods are evaluated using the five the car-following models, GM, Bando, Gipps, FRESIM and IDM, based on NGSIM (Next Generation Simulation) data. This paper proposes a framework the car-following model considering the following vehicle; the most existing car-following models can be fit into this framework. On the basis of the modeling framework, Chandler, GM, Bando, Jiang, Gipps and NETSIM models are transformed from the existing form to a form including the following vehicle. The six new models are calibrated using the improved calibration method based on NGSIM data and analyzed for the different contributions of the forward control in the car-following behavior considering the following vehicle. Stability as a critical property of the car-following model is also explored using the theoretical analysis method and the pseudo ring-road simulation method. The stability of Gipps model is studied using the theoretical analysis method, while the stability of RV (Relative Velocity) and OV (Optimal Velocity) models are examined using the pseudo ring-road simulation method.The following conclusions are drawn:1. the two improvement methods of the car-following model calibration are verified to be effective by simulations;2. Different car-following models may have different optimal objective functions in calibration; however, when the optimal objective function of a model is not clear, the weighted MAE (Mean Absolute Error) with the position as the variable will be a better objective function choice which can obtain relatively high performance for most car-following models. The weighted Theil’s U function with the velocity as the variable will be the objective function that can effectively control the simulation error accumulation in most models;3. Taking into account the following vehicle in car-following models can better describe the car-following behavior in reality;4. In the existing traffic environment, the forward control contribution is generally more than80%in the entire control task, and the backward control generally has very small proportion. Most models display the characteristic that drivers are often forced to accelerate in the car-following behavior considering the following vehicle;5. With the increase of the backward control contribution, the forward and backward acceleration distributions of some proposed models becomes unreasonable, revealing that the high backward control contribution will fail car-following behavior, so the backward control contribution should be bound in a reasonable range in the vehicle control system;6. The general stability condition of the car-following model considering the following vehicle is verified to be correct, but it is only applicable to the models in which the decision variable is acceleration and the reaction time is not included, and the formula is not very complicated;7. The Gipps car-following model considering the following vehicle displays that the backward control has three kinds of effect on traffic flow:stabilizing, destabilizing, producing non-physical phenomena, which enriches the existing results of the stability of the car-following model considering the following vehicle;8. The pseudo ring-road simulation results show that free flow can evolve into a severe traffic congestion in the stability testing, and the congested traffic flow may also evolve into a free flow accompanied by a minor traffic fluctuation;9. Generally speaking, it is not necessary to consider the influence of the following vehicle on the car-following behavior in free flow, and the backward control contribution in the total decision may need to be limited in a reasonable range to avoid the failure of the traffic cooperation and control.更多还原