自动驾驶技术上路还有哪些问题？

一旦先进驾驶辅助系统(Advanced Driver Assistance Systems，ADAS)开始在消费性汽车市场扩增，新闻媒体也跟着兴奋地预测自动驾驶车辆将会问世；市场对于自动驾驶技术的讨论越热烈，也让更多消费者认为它终会实现…不过，并不会那么快。ADAS目前的功能性与消费者对于未来自动驾驶车辆的想象之间还有很明显差异；汽车制造商与消费者之间关于自动驾驶车辆的讨论，几乎还没有开始。 有一点很重要的是，工程领域──包括汽车电子、手机或是一般消费性电子──已经不再仅与科技能做到什么有关，而是关于技术的便利性，或是有的时候不经意对人 们的生活所产生的不便。消费者已经知道如何根据科技来调整他们的行为，或是像产业界所说，去“习惯”那些科技；但可别相信那是一个既定的事实。 今日有许多工程议题，是以科技最终能如何仿真、模仿，以及如果可能，如何了解人们的行动、喜好以及情绪为中心；而这些都是自动驾驶车辆所面临的挑战。因为汽车是日常工具，自动驾驶车辆可能做出的决定对大多数消费者来说都非常需要个人化，其设计关键在于预测使用者可能或不可能做什么事情，与一个人在驾驶车辆时的行为有关。 驻地在德国慕尼黑的飞思卡尔半导体(Freescale)安全性与车身底盘部门经理Davide Santo接受EETimes美国版访问时，谈到了对未来自动驾驶车辆的客观看法；根据他的观点，自动驾驶车辆要真正上路，汽车产业至少还面临三个高阶挑战，或者说是“需要研究的地方”： 1. 车辆与驾驶人何时换手？ Santo 解释，第一个挑战与车辆与驾驶人之间的决策权换手(handover)相关，也就是知道该在何时将汽车驾驶改回“手动”；当道路上的状况对自动驾驶来说处 理不易，该种换手还可能会有好几次。而一旦系统因为某个适当动作做出“换手”的决定，必须在几秒之内就执行，但车子如何会知道驾驶人有再注意路况？在那个 “换手”的非常时刻，驾驶人要是没有在注意车子周遭状况，该怎办？ 当然，车内传感器能注意驾驶人的情况，并准备好给他提醒， 告知前方十字路口有(火车、直升机、六车连环撞…之类的)事故，然后驾驶人就该掌握方向盘准备转向；但这一切必须在非常短的时间内完成。汽车如何将控制权 交回给驾驶人、驾驶人如何接手，以及要在多快的时间内，而且汽车还一直以55英里标准时速行驶，是很令人惊悚的挑战。 2. 汽车品牌怎么推广该技术？ 当越来越多汽车业者积极投入提供自动驾驶车辆，盲目地追随风潮有可能在意料之外让一个品牌面临声誉风险。对此Santo表示：“对大多数的驾驶人来说，汽车不只是与安全性有关，还有关于驾驶乐趣。” 例 如汽车大厂BMW，该品牌自诩为“终极驾驶机器(The Ultimate Driving Machine)”供应者，为驾驶人提供“纯粹的驾驶乐趣(Sheer Driving Pleasure)”；若人们就是喜爱驾驶BMW车辆的感觉，该品牌一旦提供不用再自己开的自动驾驶车款，岂不是剥夺其乐趣？安全性当然是第一优先，但毕竟人们选择某个品牌的车辆，就是因为信任与喜爱，而汽车业者该如何将这种人性与自动驾驶车辆结合，会是一个有趣的问题。 本文授权编译自EE Times，版权所有，谢绝转载 本文下一页：那隐私权问题呢？

相关阅读：
• 自动驾驶还早，让我们先实现自动停车吧！
• 机器人已可替代人类做危险试驾测试
• 美国开放无人驾驶车测试，挂牌上路还需时日LqZesmc

{pagination} 3. 贴近驾驶人习惯的个人化功能 Santo表示，未来的汽车将会“主动”适应驾驶人：“为何我的车子不能用意大利腔说话？我的车应该要知道我就是说意大利腔英文，而且要在跟我说话时适应那样的口音。”他补充指出，这一点小小的特质，能让驾驶人的心情更放松。 今日车辆的自动座椅调整功能没人会觉得不舒服，同样的，自动驾驶车辆终将学会你的开车方法－－包括操纵方向盘、踩油门等等驾驶习惯。其底线在于自动驾驶车辆将会使用跟你一样的方法开车，这会更让你这位驾驶人跟你的新车自在相处。 那隐私权问题呢？ 上述的学习应该要在装置于自动驾驶车辆内的一个黑盒子内进行，该种盒子可纪录车子发生事故前至少三十秒的实况，也会记录你所有的驾驶习惯，包括你打方向灯的时机、以及在看到红灯时还尝试着往前移动多长距离。 “自动驾驶车辆不只该知道你开车时收了哪几封电子邮件，也得知道你会闯红灯。”Santo表示，这种监控可能会衍生出隐私权问题，特别是如果保险公司要求调阅车内纪录资料时。总之请记住，你的自动驾驶车辆就代表你本人，所以你的所有私人信息都会存在那部车子里。 根据飞思卡尔的统计，目前有75%的新车已经配备了某种程度的ADAS系统，为了降低事故率，该技术包含了行人侦测、遥控停车、在时速60公里紧急煞车以及 高速公路上自动驾驶等功能；包括该公司在内的许多芯片厂商已经准备好实现未来的ADAS技术，但汽车厂商还需要思考一些基本问题。 这些问题包括：到底智能型驾驶辅助系统要采用什么组合才恰当(例如：产业界是否会同意或是反对)？车厂将如何保护驾驶人的隐私权(例如是否允许车内资料的可移植)？以及车厂将如何针对很容易分心、与目前状况完全不同的未来驾驶人，调整其自动驾驶车辆？ 本文授权编译自EE Times，版权所有，谢绝转载 编译：Judith Cheng 参考英文原文：Straight Talk on Self-Driving Cars，by Junko Yoshida

相关阅读：
• 自动驾驶还早，让我们先实现自动停车吧！
• 机器人已可替代人类做危险试驾测试
• 美国开放无人驾驶车测试，挂牌上路还需时日LqZesmc

{pagination} Straight Talk on Self-Driving Cars Junko Yoshida MADISON, Wis. -- Once Advanced Driver Assistance Systems (ADAS) started to proliferate on the consumer automotive market, the media began to breathlessly anticipate a future of "self-driving cars." (This publication is no exception when we post a story like Can a Car Find a Parking Spot by Itself?) The louder the buzz about autonomous driving, the more excited consumers are supposed to get. They might even accept its eventuality -- or so hope auto companies. Well, not so fast. There is a pretty visible gap between what ADAS is capable of doing today and how consumers envision the ultimate future of autonomous driving. The conversation on "self-driving cars" between automakers and consumers has barely begun. It's important to note that engineering -- automotive, mobile phones, or consumer electronics in general -- is no longer just about what technologies can do. It's about how technologies convenience -- or sometimes unwittingly -- inconvenience people's lives. Consumers have been known to adjust their behavior to technologies, or "getting used to it," as the industry describes it. But don't believe for a minute that this is a given. Many engineering issues today center on how technologies eventually emulate; mimic; and, if possible, understand people's actions, tastes, and emotions. Therein lies the challenge of the self-driving car. Because a car is an everyday tool, the likely decisions made by self-driving cars must be deeply personal to most consumers. The essential design issue lies in anticipating what users might or might not do in advance. It's about knowing how a person behaves while driving. In a recent interview with EE Times, Davide Santo, Freescale Semiconductor's safety and chassis segment manager, based in Munich, offered a no-nonsense straight talk on the future of self-driving cars. In his view, there are at least three high-level challenges, or "areas for investigation," that the automotive industry needs to work on before coming up with acceptable driverless cars: Handover The first challenge is about the "handover" from car to driver -- in terms of making decisions and knowing when to return the car to "manual" control, explained Santo. There will be times, Santo explained, when the situation on the road gets too tricky for a driverless car to handle. But at a time when that "handover" decision, followed by an appropriate action, needs to be taken within a matter of seconds, how does the car know if a driver is paying attention to the road? Is the human driver aware of what's around his car at that very moment? Sure, sensors inside the car can keep an eye on the driver and be ready to get his attention, and tell him there's a train/helicopter/six-car crash brewing at the intersection, so maybe he should grab the wheel and steer for the ditch. But this all needs to happen very quickly. How the car gives back control to a driver, how the driver accepts it, and how fast -- all while the car is tooling along at 55 -- is a scary challenge. Branding As much as the automotive industry is eager to offer driverless cars, blindly following the trend could unexpectedly risk the reputation of a brand. Think about it, said Santo. "For a lot of drivers, cars aren't just about safety. They're also about the joy of driving." An automotive brand like BMW, which billed itself as "The Ultimate Driving Machine," is now offering drivers, in its promotions, "Sheer Driving Pleasure." But if a driver loves driving his BMW, how does BMW enable him to express that love in a driverless BMW that doesn't let him drive? Safety should certainly come first. But after all, people choose a car with a certain brand because they think the brand speaks to them. So, how carmakers incorporate that personality in driverless cars is an interesting quandary. Personalization Cars will adapt "actively" to drivers, according to Santo. Santo said during his EET interview: "Why shouldn't my car speak with an Italian accent? The car will know that I speak English with an Italian accent, and the car can and should adapt to it when it speaks to me." That little idiosyncrasy could easily put a driver in a much more relaxed mood, Santo added. Today, nobody is annoyed when a car does a seat adjustment for him. In the same vein, a self-driving car will eventually learn how you drive -- everything from steering to acceleration and other driving habits. The bottom line is that a driverless car will drive a car like you do. It should make you, the driver, more at home with your new car. But what about privacy? That sort of learning will be done by a black box installed in your driverless car. Just as it records the last 30 seconds of what happened before your car got into an accident, the black box will register everything about your driving habits -- from timing your turn signals to how you tend to creep forward at red lights. "The car will not only know which emails you read while driving, but also capture how you run a red light," Santo. This sort of monitoring could turn into a privacy issue, especially if insurance companies get access to the data accumulated inside each car. Remember, your driverless car will be you, and your privacy is also stored in your car. Today, 75 percent of new cars sold already incorporate some level of ADAS. With an eye towards reducing fatalities, the ADAS technology being developed includes pedestrian detection, remote parking, emergency braking at 60 km per hour, and automated highway driving, according to Freescale. The US chip company is offering safety fundamentals ranging from safe system basic chips (SBCs) to advanced 77GHz radar sensors and large, scalable MCU platforms covering front, rear, surround sensing, and fusion. Building blocks are here to enable ADAS technologies of the future. Still, automotive OEMs are grappling with basic questions, such as what should be the right set of intelligent assist systems cars (i.e., Will the industry agree or compete on this?); how OEMs plan to protect driver privacy (i.e., Will they allow transportability of data collected inside a car?); and how OEMs adjust their driverless cars to the future generation of easily distracted drivers, who are likely to be opposite of today's driver model.