交通拥堵辅助与自动化

Traffic-Jam Assistance and Automation

交通拥堵辅助与自动化

Abstract

Traffic jams are situations where a high degree of automation could give a large benefit to customers. In addition, the relatively simple situation of a traffic jam means that a high degree of automation can be expected in the near future. This chapter explores in detail the motivation, the conditions, and the versions of assistance and automation systems designed to assist in traffic jams. Different levels of automation for traffic-jam assistance and automation systems will be discussed, such as the design of traffic-jam assistance systems and of automated following systems. Moreover, HMI concepts of state-of-the-art systems are presented, and their implications on controllability and take-over scenarios are discussed. Finally, the legal situation and therefore marketability aspects are regarded.
交通拥堵是那些高度自动化能带来巨大好处的情况。此外，交通拥堵的相对简单情况意味着在不久的将来可以期待实现高度自动化。本章详细探讨了设计用于协助交通拥堵的辅助和自动化系统的动机、条件和版本。将讨论交通拥堵辅助和自动化系统的不同自动化级别，例如交通拥堵辅助系统和自动跟随系统的设计。此外，介绍了最先进系统的HMI（人机交互）概念，并讨论了它们对可控性和接管场景的影响。最后，考虑了法律状况和因此的市场推广方面。

1 Introduction

There isn’t a driver who does not find traffic jams irritating and annoying. They usually occur unexpectedly. The additional time that they consume on the way to work, to shopping, to friends, or on vacation causes a great deal of dissatisfaction, stress, and aggression.
没有一个驾驶员不觉得交通拥堵令人烦恼和恼火。它们通常出乎意料地发生。它们在上班路上、去购物、去朋友家或度假时消耗的额外时间，引起了大量的不满、压力和攻击性。
Traffic jams are the kind of situations where a high degree of automation could confer a great deal of benefit on customers when a situation cannot be avoided. Moreover, the relatively simple situation of a traffic jam means that a high degree of automation can be expected in the near future. The following will explore in detail the motivation, the conditions, and the versions of assistance and automation systems that are designed to deal with traffic jams.
交通拥堵是那种在无法避免的情况下，高度自动化能够给客户带来巨大好处的情况。此外，交通拥堵的相对简单性意味着在不久的将来可以预期实现高度自动化。接下来将详细探讨旨在应对交通拥堵的辅助和自动化系统的动机、条件和版本。

1.1 Motivation

Today’s assistance systems for stop-and-go traffic assume control only over longitudinal motion and, thus, facilitate a driver’s task only partially. By adding assistance or automation features to deal with lateral motion, it will be possible to facilitate a driver’s task still further. Counterbalancing a high degree of automation are legal questions, the cost of systems, questions of liability, and additional risks posed by automation.
当今的辅助系统在应对走走停停的交通时，只控制纵向运动，因此，只是部分地减轻了驾驶员的任务。通过增加处理横向运动的辅助或自动化功能，将进一步减轻驾驶员的任务。与高度自动化相抗衡的是法律问题、系统成本、责任问题以及自动化带来的额外风险。
Adaptive Cruise Control (ACC) systems – mostly radar assisted – that assume control over longitudinal motion have established themselves to the extent that nearly every carmaker now offers them. Recent years have seen the increasing introduction of Full Speed Range Adaptive Cruise Control (FSRA) systems (see chapter “▶ Adaptive Cruise Control”) that govern a car’s speed by bringing it to a full stop and even setting it into motion again.
自适应巡航控制系统（ACC）——主要是雷达辅助的——已经发展到几乎每个汽车制造商都提供这种系统的程度。近年来，全速范围自适应巡航控制系统（FSRA）的引入越来越多（见章节“▶ 自适应巡航控制”），它们通过使汽车完全停止甚至再次启动来控制汽车的速度。
The enhanced benefits of such systems mean that customers will probably make more use of them. Increased use of such regulating systems in turn magnifies the benefits of flowing traffic such as less CO2 emission.
这类系统的增强优势意味着客户可能会更多地使用它们。反过来，这种调节系统的增加使用又放大了流畅交通的好处，例如减少二氧化碳排放。

1.2 Acceptance

Due to their increased prevalence, a growing number of users have become acquainted with the way in which ACC systems assume control of longitudinal motion. However, drivers must permanently pay attention to lateral motion when they are stuck in traffic or in stop-and-go traffic. Studies have shown that most drivers would therefore prefer systems that could assume control over lateral motion in stop-and-go situations. In the forefront of customer expectations are relief from monotonous tasks in traffic and also the enhanced sense of safety while possibly performing other activities (Schaller et al. 2008). One can therefore assume that expanding the pure FSRA solution to include a lateral motion function will heighten the attractiveness and the chances of success of these systems.
由于其日益普及，越来越多的用户已经熟悉了ACC系统如何控制纵向运动的方式。然而，当驾驶员遇到交通堵塞或走走停停的交通时，他们必须始终关注横向运动。研究表明，因此大多数驾驶员更倾向于能够在走走停停情况下控制横向运动的系统。在客户期望的前列是减轻交通中的单调任务，并且在可能进行其他活动时增强安全感（Schaller等人，2008年）。因此，可以假设将纯粹的FSRA解决方案扩展以包含横向运动功能，将提高这些系统的吸引力和成功机会。

1.3 Definitions

In order to categorize different forms of systems designed to assist drivers in traffic jams both technically and legally, it makes sense to consider them not individually but to assess them by means of a more generalized category. A report by the German Federal Highway Research Institute (BASt) contains such a catalogue of general assistance or automated systems arranged according to categories (Gasser et al. 2012). Chapter “▶ Legal Aspects for the Development of Driver Assistance Systems” of this book details the legal aspects. Of importance for the discussion here are the definitions of the various design levels and their implications for a driver’s responsibility for driving in a traffic jam.
为了在技术和法律上对设计用来帮助驾驶员应对交通拥堵的不同系统进行分类，考虑它们不是单独的个体而是通过更一般的类别来评估它们是有意义的。德国联邦公路研究所（BASt）的一份报告包含了这样一个按类别排列的通用辅助或自动化系统的目录（Gasser等人，2012年）。本书的“▶ 驾驶辅助系统开发中的法律方面”章节详细讨论了法律方面。对于这里的讨论重要的是，各种设计级别的定义及其对驾驶员在交通拥堵中驾驶责任的影响。
Assistance systems assume control, within certain limits, over either the longitudinal or the lateral motion of a car. This presupposes that the driver monitors the system at all times and is prepared to reassume responsibility for driving. Partially automated systems assume control of longitudinal and lateral motion in certain scenarios; the driver, however, must constantly monitor the system here as well and be prepared to reassume control immediately. One way of achieving this is to require hands-on driving. Hands-off recognition (i.e., off the steering wheel) coupled with a deactivation strategy should prevent drivers from turning their attention away from traffic for longer periods of time. A major step is the transition to highly automated systems that allow drivers extra time to resume driving responsibility in specific situations, thus freeing them from the need to constantly monitor the system. The highest level of development is fully automated systems that completely free drivers from the need to monitor the system in defined circumstances.
辅助系统在一定限度内接管汽车的纵向或横向运动的控制。这假设驾驶员始终监控系统，并准备重新承担驾驶责任。部分自动化系统在特定场景中控制纵向和横向运动；然而，驾驶员在这里也必须持续监控系统，并准备立即重新控制。实现这一点的一种方式是要求手扶驾驶。手离识别（即，离开方向盘）结合停用策略应该防止驾驶员长时间将注意力从交通上转移开。一个重大步骤是过渡到高度自动化系统，这些系统允许驾驶员在特定情况下有额外的时间来恢复驾驶责任，从而免除他们持续监控系统的需要。最高级别的发展是完全自动化系统，在定义的情况下完全免除驾驶员监控系统的需要。
A practical aspect is the question as to whether, and how long, drivers should be able to leave their hands off the steering wheel. Even partially automated systems permit this already as long as the – by definition constantly vigilant – driver is able to handle unexpected situations (Gasser et al. 2012). Naturally, it would be much more convenient if drivers were able to remove their hands from the steering wheel in stop-and-go traffic for longer periods of time and devote their attention to other activities.
一个实际问题是，驾驶员是否应该能够放开方向盘，以及能够放开多长时间。即使是部分自动化系统，只要驾驶员——根据定义始终警惕——能够处理意外情况（Gasser等人，2012年），就已经允许这样做了。当然，如果驾驶员能够在走走停停的交通中长时间地将手从方向盘上移开，并将注意力投入到其他活动中，那将会更加方便。

2 Information About a Car’s Immediate Environment

Before dealing with the different types of assistance and automated systems and going into their technical aspects, we wish to first tackle the question as to what information about the surroundings should be made available to the system and how should it be collected.
在讨论不同类型的辅助和自动化系统及其技术细节之前，我们希望首先解决应该向系统提供哪些关于周围环境的信息，以及如何收集这些信息的问题。
Information on lane structures such as markings provide the basis for any traffic-jam assistance function on the one hand and information on other vehicles in the immediate environment on the other hand. Camera-based systems (cf. chapter “▶ Fundamentals of Machine Vision”) usually detect lane markings. Radar systems (see chapter “▶ Automotive Radar”) and/or camera systems with objectrecognition algorithms (see chapter “▶ Environment Representation for ADAS”) provide data on the position and movements of vehicles in the vicinity, among other things.
车道结构信息，如标线，一方面为交通拥堵辅助功能提供了基础，另一方面提供了关于周围其他车辆的信息。基于摄像头的系统（参见章节“▶ 机器视觉基础”）通常用于检测车道标线。雷达系统（见章节“▶ 汽车雷达”）和/或配备物体识别算法的摄像头系统（见章节“▶ ADAS的环境表示”）提供了关于附近车辆的位置和运动的数据等信息。
Knowledge of the lane markings and the position and motion of the vehicle immediately ahead are initially enough to provide basic traffic-jam assistance. Thus, many of today’s cars equipped with front radar for ACC systems (see chapter “▶ Adaptive Cruise Control”) and cameras for Lane Departure Warning or Lane Keeping Support (see chapter “▶ Lateral Guidance Assistance”) already possess the basic sensors needed. Any system governing lateral motion must orient itself to both lane markings and the vehicle in front because the vehicle in front may cover lane markings wholly or in part, especially in heavy traffic (Fig. 1).
了解车道标线以及紧前方车辆的位置和运动，最初足以提供基本的交通拥堵辅助。因此，许多今天配备有前雷达用于自适应巡航控制系统（ACC）的汽车（见章节“▶ 自适应巡航控制”）以及用于车道偏离警告或车道保持支持的摄像头（见章节“▶ 横向引导辅助”）已经拥有了所需的基本传感器。任何控制横向运动的系统都必须同时参照车道标线和前方车辆，因为在交通拥堵时，前方车辆可能完全或部分遮挡车道标线（见图1）。

Information on traffic to the sides and to the rear of a vehicle is indispensable for providing more complete assistance to a driver. Side-mounted and rearview cameras (surround-view systems) could monitor cars traveling alongside or lane markings alongside a car. Short- or mid-range radars could keep track of vehicles traveling alongside. Due to the relatively low speeds encountered in heavy traffic scenarios, it is possible to dispense with rear-mounted short- or mid-range radar.
为了向驾驶员提供更全面的辅助，了解车辆两侧和后方的交通信息是必不可少的。侧装和后视摄像头（环视系统）可以监控与车辆并行行驶的汽车或车辆旁边的车道标线。短程或中程雷达可以追踪与车辆并行行驶的车辆。由于在交通拥堵场景中遇到的相对低速，可以省略安装在车辆后方的短程或中程雷达。
GPS could be used at lower levels of automation to limit functions to safe scenarios (such as highway driving) that conform to predetermined areas of application. Map data could also be useful to improve system availability, however. If the complexity of the system under review grows to include highly automated functions, the inclusion of GPS positioning data and highly accurate mapping data may become necessary to provide the redundancy demanded by the number and curvature of lane markings. This information plays a decisive role in applying functions to urban scenarios in particular. Comparisons with landmarks captured above and beyond the road surface, such as bridge pillars or traffic signs, are a logical extension of such systems.
在较低级别的自动化中，可以使用GPS将功能限制在符合预定应用区域的安全场景（如高速公路驾驶）。然而，地图数据也可以用于提高系统的可用性。如果正在审查的系统的复杂性增加到包括高度自动化的功能，可能就需要包括GPS定位数据和高度精确的地图数据，以提供由车道标线的数目和曲率所要求的冗余。这些信息在特别应用于城市场景时起着决定性作用。与道路表面以上的地标进行比较，如桥墩或交通标志，是这些系统的逻辑扩展。

3 System Design Variants

3.1 Stop-and-Go Assistant with Longitudinal Regulation Only

FSRA can be considered a functional and technical basis for higher forms of traffic-jam assistants and automation (see Fig. 2), because it covers all the longitudinal regulation of a vehicle.
FSRA（全速范围自适应巡航控制）可以被视为更高形式的交通拥堵辅助和自动化的功能和技术基础（见图2），因为它涵盖了车辆的所有纵向调节。

Although FSRA provides longitudinal guidance, a driver must permanently perform lateral guidance him-/herself. The system thus remains technically and legally an assistance system by which a driver never relinquishes responsibility.
尽管FSRA提供了纵向引导，驾驶员必须始终自行执行横向引导。因此，从技术和法律上讲，该系统仍然是一个辅助系统，驾驶员永远不会放弃责任。

3.2 Traffic-Jam Assistant (Monitors Car in Front and Lane-Holding Assistant)

In contrast to the longitudinal guidance provided by an FSRA assistance system, traffic-jam assistants provide lateral guidance as well. They do so by generating steering force for the purpose of keeping a car on a predefined path.
与FSRA辅助系统提供的纵向引导不同，交通拥堵辅助系统也提供横向引导。它们通过生成转向力来实现保持汽车在预定义路径上。
Two different kinds of information go into the calculation of such an intended trajectory as detailed in Sect. 2. One concerns the position and, thus, the movement of nearby vehicles, while the other concerns lane markings. In order for path controlling to work, the intended trajectory must consist of data on deviation y0 (eccentricity to the middle of the lane), orientation yD (heading) relative to the trajectory, and its curvature kT (cf. Fig. 3). As this is the input for lateral guidance as shown in Fig. 4, these values should be observed through a Kalman filter. The following illustrates how to design this filter.
如第2节详细所述，计算预期轨迹需要两种不同类型的信息。一种涉及附近车辆的位置和运动，另一种涉及车道标线。为了使路径控制有效，预期轨迹必须包含偏离$y_0$（相对于车道中心的偏心率）、相对于轨迹的朝向 ${\theta }_{\Delta }$（方向）以及曲率$k_T$的数据（参见图3）。由于这是图4所示横向引导的输入，这些值应该通过卡尔曼滤波器来观察。以下内容说明了如何设计这个滤波器。


The following equations can serve to illustrate the movement of a vehicle along a trajectory, taking into account the vehicle’s own movement (vehicle speed vEV, slip angle bEV, and yaw rate C_ EV ):
以下方程可以用来说明车辆沿轨迹的运动，同时考虑到车辆自身的运动（车辆速度 $v_{EV}$，滑移角${\beta }_{EV}$，和偏航率 ${\dot{\Psi }}_{EV}$)：

In order to take the relative position of the vehicle in front PFC(xFC/yFC) into account, we assume that the vehicle in front will follow the intended trajectory, whereby xT ¼ xFC and yT ¼ yFC:
为了考虑前面车辆的相对位置 $P_{FC}(x_{FC}/y_{FC})$，我们假设前面的车辆将遵循预期轨迹，其中 $x_T=x_{FC}和y_T=y_{FC}$：

Camera information on the relative position of lane markings and, thus, the distance to the middle of the lane, y0_BV, the orientation yD_BV to the lines, and the curvature of the lane kT BV can serve directly to compute a lateral target trajectory to the middle of the lane (the midpoint between two lane markings). The BV index illustrates that the information on lane markings comes from image processing.
摄像头信息关于车道标线的相对位置，因此，到车道中心的距离 $y_{0\_BV}$，相对于车道线的朝向${\theta }_{\Delta \_BV}$，以及车道的曲率 $k_{T_{BV}}$，可以直接用来计算到车道中心（两个车道标线之间的中点）的横向目标轨迹。BV索引表明，车道标线的信息来自图像处理。
This results in the equation:
这导致得出以下方程：

Depending on the expected quality of the information going into the calculation of the target trajectory, the covariance matrix Q can serve to achieve a weighting between the vehicle ahead and lane markings.
根据预期输入目标轨迹计算的信息质量，协方差矩阵Q可以用来实现前面车辆与车道标线之间的加权。
The vector calculated and, hence, the relative target trajectory can then be used to trigger a lateral adjustment as shown in Fig. 4. This is how the system calculates a steering adjustment, which steers the car laterally back to the target line.
计算出的向量，因此，相对目标轨迹随后可以用来触发横向调整，如图4所示。这就是系统如何计算转向调整，将汽车横向引导回到目标线上。
First-generation traffic-jam assistance systems utilize information on the relative position of the vehicle ahead in cases in which sensors cannot adequately detect lane markings. The proximity of other vehicles that is typical at traffic-jam speeds often obscures the vision of ADAS cameras usually mounted behind
the windshield. At the same time, however, information about the vehicle ahead is much more available and much more stable, especially in heavy traffic situations. The traffic-jam assistant monitors the vehicle in front (see Fig. 5). The system behaves in a manner that is totally understandable for a driver: it follows the car in front. Should the car in front change lanes, the driver would have to resume complete control of lateral movement because the assistant in this version would cause his/her car to follow the other car into the new lane. The information gleaned from detection of lane markings (eccentricity to the middle of the lane, orientation to the trajectory, and curvature of the trajectory) serves, if available, on the one hand, to improve quality of regulation by taking account of global orientation and curvature information and, on the other hand, to generate a prompt command to resume control and switch off lateral assistance in case the car in front crosses the lane markings to change lanes. In this situation, it is better to follow along the lane markings (in the middle of the lane) than to simply switch off lateral assistance, as long as the markings are clearly recognized long enough in advance. The purpose of the assistance here is to keep a car in lane until a driver resumes control.
第一代交通拥堵辅助系统在传感器无法充分检测到车道标线的情况下，利用关于前方车辆相对位置的信息。在交通拥堵速度下典型的其他车辆的接近度常常遮挡了通常安装在挡风玻璃后的ADAS摄像头的视线。然而，与此同时，关于前方车辆的信息在交通拥堵情况下尤其更加可用且更加稳定。交通拥堵辅助系统监控前方的车辆（见图5）。系统的行为对驾驶员来说是完全可理解的：它跟随前方的车辆。如果前方的车辆变道，驾驶员将必须完全重新控制横向运动，因为在这个版本的辅助系统中，如果其他车辆变道，它会导致他的车辆跟随进入新车道。从车道标线检测中获取的信息（车道中心的偏心率、相对于轨迹的方向和轨迹的曲率）如果可用，一方面通过考虑全局方向和曲率信息来提高调节质量，另一方面在前方车辆越过车道标线变道时，生成立即接管控制的命令并关闭横向辅助。在这种情况下，只要车道标线能够提前足够清晰地被识别，沿着车道标线（在车道中间）行驶比简单地关闭横向辅助更好。这里的辅助目的是保持汽车在车道内，直到驾驶员重新控制。

From a driver’s point of view, it is worth striving for lateral assistance within a lane based on lanemarking recognition. This means that development work on traffic-jam assistants will go more in the direction of lane-holding assistance in the future. Greater availability and more stable recognition of lane markings will be necessary for this purpose. Refined camera sensors that are already in use in parkingassistance systems that contain image-processing algorithms (based on information obtained from cameras mounted on outside rearview mirrors and the rear of a car) can achieve this. Furthermore, expanded sensors on the side and rear of a car can provide information on nearby vehicles to permit a car to flow along in traffic even though lane information may not be available over the short term (Schaller et al. 2008).
从驾驶员的角度来看，值得努力实现基于车道标线识别的横向辅助。这意味着交通拥堵辅助系统的发展将更多地朝着未来车道保持辅助的方向发展。为此，将需要更高可用性和更稳定的车道标线识别。已经在停车辅助系统中使用的、包含图像处理算法的精细摄像头传感器（基于从安装在车外后视镜和汽车后部的摄像头获得的信息）可以实现这一点。此外，汽车侧面和后部扩展的传感器可以提供附近车辆的信息，允许汽车即使在短期内没有车道信息的情况下也能在交通中流畅行驶（Schaller等人，2008年）。
Although we must assume that a lane-holding assistance system (see Fig. 6) will meet a driver’s expectations, it is more of a partially automated system. The fact that it cannot be considered a highly automated system is due to limitations imposed on sensors by the possibility of false recognition or unavailability. A driver must constantly monitor a lane-holding assistant.
尽管我们必须假设车道保持辅助系统（见图6）将满足驾驶员的期望，但它更多地是一个部分自动化系统。它不能被视为高度自动化系统，这是由于传感器可能存在的错误识别或不可用性所施加的限制。驾驶员必须持续监控车道保持辅助系统。

3.3 Automatic Following in a Given Speed Range

The version of a traffic-jam assistant described above meets the expectations of users, but it does require a driver to constantly monitor the system. The value of this system thus primarily lies in its ability to relieve drivers, but they will formally not be able to devote their attention to other things. A highly automated lane-following system (see Fig. 7) that conforms to the BASt listing (see Sect. 1.3) would be the next logical step. The fact that it would permit moderate attention to things other than driving would confer palpable value on such a lane-following system.
上述描述的交通拥堵辅助系统满足了用户的期望，但它确实需要驾驶员持续监控系统。因此，这个系统的价值主要在于其能够减轻驾驶员的负担，但他们在形式上不能将注意力投入到其他事情上。符合德国联邦公路研究所（BASt）列表的高度自动化车道跟随系统（见图7）（见第1.3节）将是下一个合乎逻辑的步骤。这样的车道跟随系统将允许适度地将注意力转移到驾驶以外的事情上，这将为这样的车道跟随系统带来明显的价值。

In spite of comparable functionality, the demands on such a system are much higher than on a laneholding assistant. Since a driver is no longer required to constantly monitor the system, there must be assurance that the system does not cause a car to change lanes unintentionally. This results in greater demands on the availability of lane information.
尽管功能相似，但对这样一个系统的要求比对车道保持辅助系统要高得多。由于驾驶员不再需要持续监控系统，必须确保系统不会导致汽车无意中变道。这导致对车道信息的可用性有更高的要求。
Moreover, a driver must have sufficient time to resume driving if the system reaches its limits – in contrast to an assistance system that follows a car ahead. In the case in point, this would be the maximum speed for an automatic lane-following system, but it could also be in case the system fails. The system must be highly available in order not to negate the convenience it confers on a driver by requiring him/her to resume control too often. Combining highly precise map data with a model of the road would be one way to heighten the availability of the system if, for example, lane markings are missing. Since information on a car ahead is no long part of the equation necessary for calculating a target trajectory, the equation presented in Eq. 3 becomes simpler:
此外，如果系统达到其极限，驾驶员必须有足够的时间来恢复驾驶——与跟随前方车辆的辅助系统不同。在这种情况下，这将是自动车道跟随系统的最大速度，但也可能是系统故障的情况。为了不因要求驾驶员太频繁地恢复控制而抵消它给驾驶员带来的便利，系统必须具有很高的可用性。如果车道标线缺失，将高度精确的地图数据与道路模型结合起来将是一种提高系统可用性的方法。由于关于前方车辆的信息不再是必须的计算目标轨迹的方程的一部分，因此在方程3中呈现的方程变得更简单：

The l index indicates that the values for calculating the target trajectory may stem from different sources such as a frontal camera, parking camera, or highly accurate digital maps.
$l$ 索引表明，用于计算目标轨迹的值可能来自不同的来源，如前置摄像头、停车摄像头或高度精确的数字地图。
This places great demands on redundancy concepts for capturing surroundings, for functions, and for activating systems. The system must be designed in such a way that no critical situations arise that the car cannot handle itself within the defined take-over time.
这为捕捉周围环境、功能和激活系统的冗余概念提出了很高的要求。系统必须设计得当，以确保不会出现汽车在定义的接管时间内无法自行处理的危急情况。
The system’s speed limit should be set – as in the case of assistance systems – in such a way so as to permit a car to “flow along” in typical stop-and-go situations. At the same time, the speed should not be so high as to place unnecessary demands on the system. Studies of traffic jams have revealed that a maximum speed of 50 km/h permits drivers to comfortably avail themselves of the system (Sandk€ uhler 2002).
系统的限速应该设置得——就像辅助系统一样——允许汽车在典型的走走停停情况下“顺畅行驶”。同时，速度也不应该过高，以免给系统带来不必要的要求。对交通拥堵的研究表明，最高速度设定为50公里/小时可以让驾驶员舒适地使用该系统（Sandkühler 2002）。

4 Interaction of Driver and System

One important aspect of creating such systems – also of assistance-system versions – will be shaping the human-machine interface (HMI). A driver should have a clear understanding of the system status at all times so as to be able to react intuitively in an emergency situation. Systems for recognizing a driver’s status (see chapter “▶ Driver Condition Detection”) will permit dynamic interaction between car and driver in the future. The point in time for a driver to resume control of a car will depend on how much attention the driver is currently paying.
创建这类系统——包括辅助系统版本——的一个重要方面将是塑造人机界面（HMI）。驾驶员应始终清楚地了解系统状态，以便在紧急情况下能够直观地作出反应。识别驾驶员状态的系统（见章节“▶ 驾驶员状态检测”）将允许未来汽车与驾驶员之间的动态互动。驾驶员重新控制汽车的时间点将取决于驾驶员目前所付出的注意力多少。

4.1 Human-Machine Interface (HMI)

In contrast to a system that governs purely longitudinal motion (FSRA), adding lateral assistance places additional demands on the HMI. It represents a new mode that has to be available only in certain situations such as a traffic jam or stop-and-go traffic.
与仅控制纵向运动的系统（FSRA）相比，增加横向辅助对人机界面（HMI）提出了额外的要求。它代表了一种新模式，这种模式只能在特定情况下，如交通拥堵或走走停停的交通中使用。

4.1.1 BMW Human-Machine Interface

The following is a presentation of the BMW i3 traffic-jam assistant. The purpose of the assistant is to relieve drivers of the need to move longitudinally or laterally when stuck in traffic. The system that BMW offers needs only a mono camera to direct longitudinal and lateral movement. BMW has been offering this system at a comparatively low price in all its cars since 2013.
以下是对宝马i3交通拥堵辅助系统的介绍。该辅助系统的目的是减轻驾驶员在交通拥堵时进行纵向或横向移动的需要。宝马提供的系统仅需一个单目摄像头来控制纵向和横向运动。自2013年以来，宝马一直在其所有车型中以相对较低的价格提供这一系统。
The steering wheel and the lengthened side bars depict active longitudinal and lateral motion support in traffic here, as in Fig. 8a. If the function is activated and certain other conditions are fulfilled (e.g., map data verifies the highway), but the vehicle speed exceeds the functional range of 0–60 km/h, only the side bars appear (Fig. 8b). ACC is active; lateral guidance is in standby. If the speed drops below 60 km/h, lateral guidance kicks in without any user interaction. If a system limit is exceeded (e.g., if the system detects a hands-off situation) and the lateral guidance system is thus deactivated, an optical signal (red blinking steering wheel in Fig. 8a) and audible information for the driver are generated.
方向盘和加长的侧边条在交通中表示活跃的纵向和横向运动支持，如图8a所示。如果功能被激活并且满足其他一些条件（例如，地图数据确认是高速公路），但车速超过0-60公里/小时的功能范围，只有侧边条会出现（见图8b）。自适应巡航控制（ACC）处于激活状态；横向引导处于待命状态。如果速度降至60公里/小时以下，横向引导会在不需要用户交互的情况下启动。如果超过系统限制（例如，系统检测到手离方向盘的情况），横向引导系统因此被停用，将为驾驶员产生一个光学信号（如图8a中的红色闪烁方向盘）和可听到的信息。

4.1.2 Daimler Human-Machine Interface

Daimler introduced DISTRONIC PLUS with steering assistant and stop-and-go pilot as part of its Mercedes-Benz Intelligent Drive in 2013 in the new S-Class (Daimler 2013) and the facelifted E-Class, the first time that it offered a comprehensive driver-assistance system for longitudinal and lateral motion over the entire range of speed. At the lower speed range, the system follows a vehicle ahead and also orients itself to any lane markings it recognizes (stop-and-go pilot). As the speed increases, the system only reacts to lane markings, whereby it does take account of other traffic and any road restrictions it detects. The lateral guidance system does not switch off at boundary speeds. Instead, it transits seamlessly to a lane-centering system. Lateral guidance support for the entire range of speed works only if longitudinal guidance (DISTRONIC PLUS) has been activated. There is a button which can be used to activate it separately (see Fig. 9). A gray steering-wheel symbol appears in the instrument cluster in addition to the LED display. The steering wheel’s color changes to green as soon as lateral guidance becomes active.
戴姆勒在2013年推出了配备转向辅助和走走停停领航的DISTRONIC PLUS，作为其梅赛德斯-奔驰智能驾驶系统的一部分，首次在新S级（戴姆勒2013年）和改款E级中提供全面的驾驶员辅助系统，用于整个速度范围内的纵向和横向运动。在较低的速度范围内，系统跟随前方车辆，并且还会根据它识别到的任何车道标线进行定位（走走停停领航）。随着速度的增加，系统只对车道标线做出反应，同时它确实会考虑到它检测到的其他交通和任何道路限制。横向引导系统在边界速度时不会关闭。相反，它无缝过渡到一个车道居中系统。只有在纵向引导（DISTRONIC PLUS）被激活时，整个速度范围的横向引导支持才能工作。有一个按钮可以用来单独激活它（见图9）。除了LED显示屏外，仪表盘上还会出现一个灰色的方向盘符号。一旦横向引导激活，方向盘的颜色会变为绿色。

4.2 Handover and Controllability

Any technical system designed to guide motor vehicles laterally and longitudinally has to have a way to hand over control of the vehicle back to a driver in case of an error or a traffic situation that its specifications do not anticipate (system limits). One variable among the various versions of systems is the time that the system takes to transfer control back to a driver.
任何设计用来横向和纵向引导机动车辆的系统都必须有一种方法，在发生错误或遇到其规格无法预见的交通情况（系统限制）时，将车辆控制权交还给驾驶员。不同版本系统之间的一个变量是系统将控制权交还给驾驶员所需的时间。
The need for a handover can occur suddenly in the case of partially automated driver-assistance systems. That is the reason why these systems are designed so that a driver must be present and must constantly monitor the trip so that he/she can intervene when necessary. It is known that a high degree of trust in a system can lead to a driver reacting slower to system limits (Niederée and Vollrath 2009). Trust in the system in turn depends heavily on the perceived reliability of the system (Niederée and Vollrath 2009). Unavailability or experience with system errors, on the other hand, reduce driver acceptance. As the degree of support a system lends increases, the system will continually monitor a driver’s attention via such things as the hands-off recognition shown in Sect. 1.3.
部分自动化驾驶辅助系统在需要移交控制权时可能会出现突然情况。这就是为什么这些系统被设计成必须有驾驶员在场，并且必须持续监控行程，以便在必要时进行干预。众所周知，对系统的高度信任可能导致驾驶员对系统限制的反应变慢（Niederée和Vollrath，2009年）。反过来，对系统的信任在很大程度上取决于系统感知的可靠性（Niederée和Vollrath，2009年）。另一方面，系统的不可用性或对系统错误的体验会降低驾驶员的接受度。随着系统提供的支持程度增加，系统将通过诸如第1.3节所示的离手识别等方式不断监控驾驶员的注意力。
Highly automated systems free drivers from the need to constantly monitor traffic within a defined scenario, allowing them to take their hands off the wheel while they pursue other activities (within reason). The length of time that a driver can remove his/her hands from the wheel is a function of the extent and the quality of the scenario covered. When the system hands control of the car back to the driver, the question arises as to how quickly and how well this will occur and what influence the type of side activity will have on the handover.
高度自动化系统在定义的场景中解放了驾驶员，使他们无需持续监控交通，允许他们在进行其他活动（在合理范围内）时将手从方向盘上移开。驾驶员可以将手从方向盘上移开的时间长度取决于覆盖场景的范围和质量。当系统将汽车的控制权交还给驾驶员时，就会出现如何快速以及如何顺利地进行这一过程的问题，以及进行的副活动类型将对控制权移交产生何种影响的问题。
Zeeb and Schrauf (2014) distinguish between two aspects of the driver resuming control. (1)“Formal” resumption, including the initial intervention. It becomes possible as soon as a driver is ready in the sense of hands on the wheel and eyes fixed on the road. (2) Resumption which is adequate to avoid an accident. It involves cognitive recognition plus the choice of an adequate reaction. The time for adequate resumption depends on whether a driver must completely reorient him-/herself or whether he/she still has a valid mental model of the traffic situation and only has to update it. A study of a critical traffic situation reveals that drivers will take between 1.6 and 2.3 s once they have been requested to resume control, before hitting the brakes.
Zeeb和Schrauf（2014年）区分了驾驶员恢复控制的两个方面。(1) “形式上”的恢复，包括最初的干预。一旦驾驶员在手放在方向盘上、眼睛注视路面的意义上准备好了，这就成为可能。(2) 足以避免事故的恢复。它涉及认知识别加上选择适当的反应。适当的恢复时间取决于驾驶员是否必须完全重新定位自己，或者他/她是否仍然有一个有效的交通情况心理模型，而只需要更新它。对一个关键交通情况的研究表明，一旦驾驶员被要求恢复控制，他们会在1.6到2.3秒之间采取行动，然后才会踩刹车。
Aside from the time it takes for a driver to resume control, it is necessary to consider the quality of that resumption and the control that goes with it when it comes to highly automated driving. Experience regarding control is still thin. Studies up till now have addressed normal driving and have examined such things as disturbances in electronic power steering and their controllability (Neukum et al. 2009, 2010).
除了驾驶员恢复控制所需的时间外，在高度自动化驾驶的情况下，还需要考虑恢复的质量以及随之而来的控制。关于控制的经验仍然有限。到目前为止的研究已经涉及正常驾驶，并检查了诸如电子动力转向中的干扰及其可控性等问题（Neukum等人，2009年，2010年）。
Damböck et al. (2012) have noted no significant differences in the quality of how a situation is handled exist after a period of 6–8 s following a handover, compared to a group of normal drivers. The group examined performed intense manual visual and cognitive side activities in noncritical traffic situations. The study addressed tasks associated with stabilizing, steering, and navigating. In none of the three variations does the study represent a critical traffic situation. Instead, drivers could use all of the relatively long time made available to them. The influence of a variation in the side activity as observed by Petermann-Stock et al. (2013) led to maximum resumption lags between 2.4 and 8.8 s, regardless of the age or sex of the people tested. Eight different studies by Giesler and M€ uller (2013) brought to light similar results. The median time for resumption was 2.7 s, while the maximum was likewise 8.8 s.
Damböck等人（2012年）指出，在控制权移交后6-8秒的时间内，处理情况的质量与正常驾驶员群体相比没有显著差异。所研究的群体在非关键交通情况下进行了密集的手动视觉和认知副活动。研究涉及与稳定、转向和导航相关的任务。在三种变化中，研究都没有代表一个关键的交通情况。相反，驾驶员可以利用提供给他们的所有相对较长的时间。Petermann-Stock等人（2013年）观察到的副活动变化的影响导致了2.4到8.8秒之间的最大恢复延迟，无论被测试者的年龄或性别如何。Giesler和Müller（2013年）的八项不同研究揭示了类似的结果。恢复的中位时间为2.7秒，而最大时间同样是8.8秒。
In general, one can say that the studies of highly automated driving that have been conducted thus far have revealed time lags before resumption of up to around 9 s. In noncritical situations, therefore, the time lag was quite long, whereas in critical traffic situations, very short reaction times of around 2 s were observed. The urgency of the situation obviously exerts a great influence on the time needed to resume control, but it can be to the detriment of quality.
总的来说，可以说到目前为止进行的高度自动化驾驶研究揭示了在恢复控制之前的时滞高达约9秒。因此，在非关键情况下，时滞相当长，而在关键交通情况下，观察到的反应时间非常短，大约为2秒。显然，情况的紧迫性对恢复控制所需的时间有很大的影响，但这可能会损害控制的质量。
Also interesting is the question how a system reacts if a driver does not respond to a return of control. Deactivating the function is certainly the technically easiest solution here. This is the usual practice with partially automated functions. Depending on the situation, gradually shutting down the function is preferable to abruptly deactivating it. Whether it is better to deactivate the function after the time for resumption has expired or whether it is preferable to perform a minimal risk maneuver is still an open question which surely will depend on the scenario.
同样有趣的问题是，如果驾驶员不对控制权的返回做出响应，系统会如何反应。从技术上讲，最简单的解决方案是停用该功能。这是部分自动化功能通常的做法。根据情况，逐渐关闭功能可能比突然停用它更可取。在恢复时间到期后停用功能是否更好，或者执行最小风险操作是否更可取，这仍然是一个开放的问题，这肯定会取决于具体场景。
Studies focusing on a driver’s resuming responsibility for driving, especially in heavy traffic situations, are not known. Due to the low speed, it is not anticipated that drivers will “overreact.” Furthermore, it is no problem to go for the safe status of “stop” making it easy to tap it as a solution in critical traffic situations.
关注驾驶员在交通拥堵情况下恢复驾驶责任的研究尚不清楚。由于速度较低，预计驾驶员不会“过度反应”。此外，选择“停车”的安全状态并不成问题，这使得在关键交通情况下很容易将其作为一种解决方案。
Fully automated systems represent the highest degree of automation. They relieve drivers of the necessity of monitoring traffic for a long time. What requirements result for procedures aimed at abandoning fully automated driving, and how long such a handover really takes, still gives rise to a multitude of questions. One need only imagine the case of a system handing over control to a driver who is napping, to put the question into perspective.
完全自动化系统代表了自动化的最高程度。它们长时间解放了驾驶员监控交通的必要性。针对放弃完全自动化驾驶的程序有哪些要求，以及这样的交接实际上需要多长时间，仍然引发了许多问题。人们只需想象一下系统将控制权交给正在打盹的驾驶员的情况，就可以将问题置于正确的视角。

4.3 Marketability

4.3.1 The Legal Situation

The same legal challenges and problems of liability present themselves for traffic-jam assistants (and especially for traffic-jam automation) as they do for other assistance systems or automation scenarios. Whereas accidents caused by counterparties should not pose any problem, there is a gray zone between carmakers and customers when it comes to damage caused by technical failures or damage caused by a driver’s own fault, for example, by ignoring a command to resume control of the car.
交通拥堵辅助系统（尤其是交通拥堵自动化）与其他辅助系统或自动化场景一样，面临着相同的法律挑战和责任问题。由于对方造成的事故不应该存在任何问题，但在技术故障造成的损害或驾驶员自身过错（例如，忽视恢复对汽车控制的指令）造成的损害方面，汽车制造商和客户之间存在一个灰色地带。
These aspects must still be clarified before introducing and marketing highly automated systems for stop-and-go traffic. The highly automated traffic-jam assistance system is currently not permitted for speeds above 10 km/h (ECE 2006). Refer to chapter “▶ Legal Aspects for the Development of Driver Assistance Systems” for a more detailed discussion of these topics.
在引入和推广用于走走停停交通的高度自动化系统之前，这些方面仍需澄清。目前，高度自动化的交通拥堵辅助系统在速度超过10公里/小时时是不被允许的（ECE 2006年）。有关这些主题的更详细讨论，请参阅章节“▶ 驾驶辅助系统开发中的法律方面”。

4.3.2 Analysis of Marketability

Compared to implementing highly automated functions at high speeds, the specific automation of trafficjam scenarios has several advantages that simplify its technical and commercial feasibility greatly. Automating longitudinal and lateral movement at typical stop-and-go speeds of up to around 50 km/h requires less sensor range compared to automation at higher speeds. The system also does not rely on a complicated and expensive back end. The effort needed to ensure function in case of a failure (such as if an actuator fails) diminishes, again due to the short braking distances involved. Stop-and-go traffic on highways, in particular, is relatively easy to describe and comprehend. There are no intersections and no traffic lights (except in the case of tunnels), nor does one normally have to deal with pedestrians or cyclists. The curves on highways, moreover, exhibit a generous radius. These factors reduce the complexity of the scenario and also greatly reduce the demands on sensors and on models of the surroundings. The speeds and positions of surrounding vehicles and the lane markings are enough to regulate the function. A more wide-ranging and detailed detection of maneuvering space is, at least for the simpler versions of the system (see Sect. 3), unnecessary. Whether it will be possible to implement the technical safety concept using the components found in today’s automobiles (sensors and actuators) due to the low speed and the possible effects of a system failure is the subject of debate at the current time.
与在高速下实现高度自动化功能相比，特定交通拥堵场景的自动化具有几个优势，这些优势极大地简化了其技术可行性和商业可行性。与在更高速度下的自动化相比，在典型的走走停停速度（大约50公里/小时）下自动化纵向和横向运动需要较少的传感器范围。系统也不依赖于复杂且昂贵的后端支持。由于涉及的制动距离短，确保故障情况下功能所需的工作量也减少了。特别是高速公路上的走走停停交通，相对容易描述和理解。没有交叉路口和红绿灯（隧道情况除外），通常也不需要处理行人或骑自行车的人。此外，高速公路上的曲线半径也很大。这些因素降低了场景的复杂性，也大大降低了对传感器和周围环境模型的要求。周围车辆的速度和位置以及车道标线足以调节功能。至少对于系统的较简单版本（见第3节），不需要更广泛和详细的操纵空间检测。目前，由于速度较低以及系统故障可能产生的影响，是否可以使用当今汽车中的组件（传感器和执行器）来实现技术安全概念，这是一个有争议的话题。
If it were just a matter of one car and the ones immediately surrounding it, a stop-and-go situation on a highway is hardly distinguishable from traffic situations on other roads or in urban environments (see Fig. 1). Urban traffic, however, presents special challenges for automated systems. In addition to intersections, turn lanes, and traffic lights, there are also nonmotorized participants. A driver in urban traffic must reckon with pedestrians or cyclists moving among cars at any time. Any system that is going to provide safe automation in this kind of environment is going to have to be able to recognize nonmotorized participants reliably and at any time. The system must be able to recognize intersections and other complex scenarios early enough so that drivers have sufficient time to take over when a system reaches its limits. These points place greater demands on sensors, models of the surroundings, and analysis than a highway situation. The result is that a system designed to automate a traffic-jam situation on a highway is not necessarily transferrable to an urban environment. The sensors for a system designed for use on a highway are insufficient for recognizing reliably whether a vehicle is in urban traffic or stuck in a traffic jam on a highway. One possible solution would be to enlist the aid of map data from a common navigation system to assist in differentiating.
如果只是涉及到一辆汽车及其周围紧邻的车辆，高速公路上的走走停停情况几乎无法与其他道路上或城市环境中的交通情况区分开来（见图1）。然而，城市交通为自动化系统带来了特别的挑战。除了交叉路口、转弯车道和交通信号灯，还有非机动参与者。城市交通中的驾驶员必须随时准备应对在汽车之间穿行的行人或骑自行车的人。任何旨在在这种环境中提供安全自动化的系统都必须能够可靠地随时识别非机动参与者。系统必须能够及时识别交叉路口和其他复杂场景，以便当系统达到其极限时，驾驶员有足够的时间接管。这些要求对传感器、周围环境模型和分析提出了比高速公路情况更高的要求。结果是，为高速公路上的交通拥堵情况设计的自动化系统并不一定适用于城市环境。为高速公路设计的系统的传感器不足以可靠地识别车辆是在城市交通中还是卡在高速公路上的交通拥堵中。一种可能的解决方案是利用通用导航系统的地图数据来协助区分。
This could give rise to a market gap. Customers who have had a positive experience with longitudinal and lateral guidance in traffic situations on highways would like to see expanded functionality, such as on other roads or in an urban environment.
这可能导致市场缺口的出现。在高速公路交通情况下体验过纵向和横向引导的客户，希望能够看到功能扩展，例如在其他道路或城市环境中也能使用。

5 Final Remarks

FSRA systems are becoming more popular even in smaller cars. Systems that govern solely the longitudinal motion of a car have achieved a level of quality and availability (see chapter “▶ Adaptive Cruise Control”). Systems that govern lateral motion have now established themselves as well (see chapter “▶ Lateral Guidance Assistance”). It is now time to take the next step and combine longitudinal and lateral guidance systems into a new, comprehensive system. This could take the form of classic partial automation. However, a highly automated system that permits drivers to pursue non-driving activities would represent a really large step forward in terms of benefits and acceptance. An automatic assistant for keeping in lane while stuck in traffic at relatively low speeds could represent an initial system. The disadvantage of such a system is that it reaches its limits when a certain top speed is attained, something that can happen more or less frequently, depending on how traffic is flowing. This will quickly instill the desire among customers for a comprehensive system covering the whole range of speed. It is not yet possible to establish highly automated systems due to legal and technical reasons. Mercedes-Benz and BMW have already taken an initial step in making longitudinal and lateral assistance systems a reality in 2013 as they introduced lateral guidance functions. This is certainly just the beginning of a long way to automated driving, and the market will witness the introduction of many more systems.
FSRA系统甚至在小型汽车中也越来越受欢迎。仅控制汽车纵向运动的系统已达到一定的质量和可用性水平（见章节“▶ 自适应巡航控制”）。现在，控制横向运动的系统也已经确立了自己的地位（见章节“▶ 横向引导辅助”）。现在是时候迈出下一步，将纵向和横向引导系统集成到一个全新的综合系统中。这可以采取经典部分自动化的形式。然而，一个允许驾驶员从事非驾驶活动的高级自动化系统将在效益和接受度方面代表一个真正的大步前进。一个在相对较低速度的交通拥堵中保持车道的自动助手可以代表一个初始系统。这样的系统的缺点是，当达到一定的最高速度时，它就会达到极限，这种情况发生的频率可能因交通流量而或多或少。这将很快在客户中激发出对覆盖全速度范围的全面系统的需求。由于法律和技术原因，目前还不能建立高度自动化系统。梅赛德斯-奔驰和宝马已经在2013年通过引入横向引导功能，为实现纵向和横向辅助系统迈出了第一步。这当然只是通往自动化驾驶漫长道路的开始，市场将见证更多系统的推出。