人们常说，沟通是建立信任的关键。人与人之间需要坦率沟通，才能建立彼此的信任。然而，无论对于何种人际关系而言，如何沟通都是一件十分复杂的问题，各方往往无法就同一事件达成共识。

而现在，如果把关系的当事方从“人与人”变为“人与自动驾驶汽车”，想必要达到“心有灵犀”，难度只会有增无减。而这也正是摆在日产的Melissa Cefkin博士及其同事面前的难题。不夸张地说，对于业内研发自动驾驶汽车的所有公司而言，这都将是迈向下一个出行时代所必须面对的挑战。

尽管一直以来我们都想当然地认为汽车研发应该就是工程师和设计师份内的事，然而随着自动驾驶汽车这一概念横空出世，所需的专业知识便涉及到了各行各业。律师、伦理学家、数据专家，甚至是像Cefkin博士这样的人类学家也参与到了其中。作为日产硅谷研究中心的首席科学家，以及设计项目中的人类学家，Cefkin博士正带领研究小组，积极研究人们是如何看待无人驾驶汽车的，而在未来都市中人类又要如何与之共存。

镜头对准繁忙的都市地铁站，只见人流行色匆匆，摩肩接踵。然而即便如此，我们还是会变换角度，观察行人四下打量的眼神，捕捉他们脸上的微表情，记录下他们活动肩膀的小动作。就像在车来车往的街上过马路时，有时行人会同驾驶员交换一个眼神，可能只是一刹那，却足以让他们判断出是可以安全穿过，还是需要站在原地，等待汽车驶过。

人类对于一些细枝末节往往会出人意料地敏感，也正是这样一种敏感，才让我们这个世界可以正常运转。自打出生起，我们就在不断学习这种能力，即便这种感知能力并非完美无瑕。无论车里有没有人驾驶，汽车都需要对其他车辆和行人及时提供信息反馈，表明自己的意图，同时也从周边环境中获得必要信息。

就我们目前所知，每一次人们要把新科技加入到现有环境中，都会面临一系列的挑战和困难。在曼哈顿的街道上随便逛逛，马路上以及行人之间眼神与声音交流所发出的大量信号，就会如狂风暴雨般扑面而来。而在自动驾驶汽车的世界里，声音和图像的变化同样惊人。各种信号和信息蜂拥而至，人们很容易就会不堪重负。

 “人们必须尽快适应这些让人眼花缭乱的视觉提示信息。”Cefkin说道，“行人要在短时间内作出判断：这辆车是否‘看见’了他/她，从而建立起必要相互信任。”

举例而言，日产和丰田等多家公司已经做出了相关的概念车，利用多种外部展示技术，很好地提供这些重要信息。日产的IDS概念车在挡风玻璃处设置了一块电子指示牌，以便让其他车辆或行人看到。然而，如果每一家制造商都各自开发自己的反馈信息系统，那么对于行人而言，要解读自动驾驶汽车的动向将会变得无比困难。

谈到反馈信息这一问题时，Cefkin表示，“我个人支持研发行业统一的系统。”尽管就规格标准已经开展了一些初步的讨论，很多问题依旧悬而未决。

另外，Cefkin一直强调的方式就是位移提示。就目前人类感知能力的局限性而言，“最容易被察觉的无疑就是汽车的移动。”例如，人们可以觉察到汽车在加速方面的变化，从而对其下一步动向保持警觉。

Cefkin加入日产的两年来，围绕自动驾驶技术的主要工作大都集中于研发感应、定位以及控制等核心技术，然而Cefkin在信息反馈方面的研究，对于整体的部署同样重要。

Cefkin坦言，一旦出错，人与自动驾驶汽车之间的沟通效果就会非常之差，其结果“可能会导致双方之间极度不信任、关系极为紧张”，甚至有可能在自动驾驶汽车还未站稳脚跟之时，就将其扼杀在摇篮之中。

目前尽管还没有透露自动驾驶汽车的研发总投入，但具体数额肯定已达到数十亿美元，并仍在增加。因此这个结果应该是大家都不想看到的。

It’s said that communication is the key to trust. Between us humans, we need to honestly communicate our intentions to build trust. But in any relationship, communicating can be a complex problem and parties often don’t draw the same conclusions from the same message.

Now shift the relationship from person-to-person to person-to-automated vehicle with “intentions” that may be far more opaque. That’s the challenge facing Dr. Melissa Cefkin and her colleagues at Nissan—and every other company involved in autonomous-vehicle development—as we move into the next era of mobility.

Although we traditionally think of vehicle development as the province of engineers and designers, the promise of automated driving has drawn a plethora of new skillsets into the process. Along with lawyers, ethicists and data scientists, there are anthropologists like Dr. Cefkin. As a principal scientist and design anthropologist at Nissan’s Silicon Valley research center, Cefkin and her team are studying how people perceive vehicles that don’t have human drivers and how they will coexist in future cities.

Traversing a busy subway station, throngs of people seem to move seamlessly, but there are constant glances around, microexpressions detected on faces or movements of a shoulder as someone slips through the crowd. When you cross a busy street, you may exchange a quick glance with a driver and that’s all it takes to judge whether it’s safe to go—or whether it’s better to wait and let the car pass.

Humans are remarkably adaptable and sensitive to nuances that make society work. It’s something we learn as we grow from infancy. It’s far from a flawless process, though. Regardless of whether someone is riding in the vehicle, these machines will have to provide feedback to other vehicles and pedestrians about their intentions and in turn read signals from other entities in the driving environment.

As we hopefully have learned, adding technology to any ecosystem typically adds a range of new challenges and problems. A stroll down a Manhattan street bombards people with sights and sounds of traffic and personal interactions. In a world of autonomous electric vehicles, the sound and visualscape changes dramatically. People could be easily overwhelmed by these new stimuli.

“People will have to adapt and change to the new visual cues they must interpret,” said Cefkin. “They need to understand quickly if the car has seen me, in order to build the necessary trust.”

Nissan and Toyota, to name a couple, have shown concept automated vehicles that leverage a variety of interesting external-display techniques designed to provide these messages. Nissan’s IDS Concept has a digital signboard in the windshield that displays messages to other road users. However, if every manufacturer goes its own way with these feedback systems, it will make it exponentially more difficult for pedestrians to interpret an automated vehicle’s likely behavior.

“I’m personally committed to developing harmonization,” Cefkin added in regard to these signals. While preliminary discussions have begun on standards, it’s still premature to lock down much of anything.

Another approach Cefkin highlights is motion cues. For all the perception limitations humans have, it appears “the most expressive thing about vehicles is their motion.” People can detect changes in acceleration, for example, that give clues to intent.

In the two years since Cefkin joined Nissan, much of the public effort around automation has been directed at developing the core technologies of perception, mapping and control, but her feedback-stimuli efforts are equally important to the deployment process.

Done wrong, the results of poor communication between people and automated vehicles “could be most profound with mistrust and discomfort” that kills adoption before it can really take hold, Cefkin warns.

With the untold billions of dollars invested—and still to be invested—in autonomous-vehicle development, I doubt anybody wants that.

Author: Sam Abuelsamid
Source: SAE Automotive Engineering Magazine