自动驾驶汽车前路漫漫而修远兮

你认为到2030年，全球将会有多少自动驾驶车辆在路上跑？在被Google的自动驾驶汽车迷倒之前，让我们用相对客观的数字来评估这种新技术吧！ 事实证明，产业专家对于自动驾驶车辆市场的看法及预测意见分歧；分析师对于2030到2035年之间的自动驾驶车辆数量预测，从2,000、3,000万台到9,500万台都有。这些不同的预测取决于他们对自动驾驶车辆之“自动驾驶”程度的定义，以及他们所预期在自动驾驶车辆市场起飞之前，将面临的法令以及商业障碍。 美 国一家市场研究机构Navigant Research对于自动驾驶车辆市场的看法也许是其中最乐观的，该公司近期发表的一份报告预测，自动驾驶车辆年销售量将由2020年的低于8,000 台，在2035年增加至9,540万台；换句话说，2020年自动驾驶车辆在全球汽车市场的渗透率约仅0.01%，但在2035年可提升至75%。 Navigant Research 对“自动驾驶”的定义与Google的无人驾驶车辆相同，也就是车辆的自动功能基本上可替代人类司机；该机构资深分析师David Alexander表示：“我们所谈的并非只有配备部分自动驾驶功能、仅能在有限情况下自动驾驶之车辆。” 例如即将上市的 2014年式奔驰S-Class车款，虽配备了也被归类为自动驾驶的高速公路巡航(freeway cruising)功能，有产业媒体指出，该车款仅能算是“非常接近自动驾驶车辆”。 另一家市场研究机构Strategy Analytics则提出了配备多种先进安全技术的车辆，以及一台真正自动驾驶车辆之间的差异；该机构所定义的自动驾驶车辆是“高度自动化(但不一定要支持完全自主驾驶)”，估计该类车辆在全球汽车市场的占有率，在2025~2030年之间为15~20%，也就是2,000万至3,000万台。 Strategy Analytics汽车市场分析师Ian Riches表示，他们认为的自动驾驶车辆是能“在多种不同的驾驶情境下，为驾驶人提供有力的支持”；但他也指出，真正的自动驾驶车辆──“也就是那种你可以坐在驾驶座上下指令，就会直接送你去电影院、机场或度假地点的”──恐怕渗透率仅会是非常小的个位数。 Riches认为，完全自动驾驶车辆：“将大概会在2025年以后才真正问世，而且仅能在一些特定情境下才能提供完全自动驾驶功能，例如在高速公路、或是具备特定程度V2X技术支持的一些区域。” 对“自动驾驶”的定义 如同Riches所言，今日市场上并没有“100%挂保证”的自动驾驶车辆定义；但美国运输部旗下的国道安全局(National Highway Traffic Safety Administration，NHTSA)倒是在今年稍早，对车辆的自动化程度定义出五个等级： 本文授权编译自EE Times，版权所有，谢绝转载 第2页：对“自动驾驶”的5个等级定义 第3页：各国法律都规定，车辆必须有驾驶人操控、且全时进行控制

相关阅读：
• 自动驾驶技术上路还有哪些问题？
• 自动驾驶还早，让我们先实现自动停车吧！
• 美国开放无人驾驶车测试，挂牌上路还需时日gcAesmc

{pagination} ˙Level 0 ──根据NHTSA的定义，此等级车辆不具备自动化功能；驾驶人须全时间完全独力操作车辆的各种主要控制功能，包括煞车、方向盘、油门与动力。 ˙Level 1 ──具备1种以上自动化控制功能的车辆，包括电子车身稳定控制、预充电煞车(pre-charged brakes)，也就是车辆会自动在煞车时提供协助，让驾驶人在紧急情况时能重新取得车辆的控制权，或者是将车辆更快煞停。 ˙Level 2 ──能同时提供组合式的自动化功能，提供至少两种可共同运作的主要控制功能，以减经驾驶人负担；例如自适应巡航控制与车道对准功能的结合。 ˙Level 3 ──提供有限的自动驾驶功能，驾驶人能在特定的路况环境下完全交出安全关键驾驶功能，而驾驶人也能在充分的交接时间下进行偶尔的车辆控制。NHTSA 就将Google的自动驾驶车辆视为这类“具备有限自动化驾驶功能”的车辆。 ˙Level 4 ──NHTSA将此等级车辆视为“完全可自动驾驶车辆”；这类车辆的设计旨在执行完整的安全关键驾驶功能，并可在车辆行驶全程中监测道路状况。这种设计预期驾驶人会提供路径指向或是导航信息输入，但不预期驾驶人会在任何时间取得车辆控制权。 对 此Strategy Analytics 的Riches指出：“我会将NHTSA所定义的Level 2、level 3车辆视为“半自动驾驶”，而Level 4车辆则可保留“完全自动驾驶”的称号；基本上，如果驾驶人需要出现在车上，或是在需要时取得车辆控制权，这种车辆仅能算是“半自动驾驶”。” 而对任何一家市场研究机构来说，预测20年以后的未来其实是一项棘手任务且为不精确的科学；Riches形容，任何延展至2025年以后的预测，其实比较偏向是“预言”，而非透过模型化分析的数据。 Navigant Research的Alexander 也坦承有这样的挑战：“20年是很长的时间，不过我们是根据现有的技术来最预测；”他表示该机构也会收集产业界正在推动的计划等信息。 本文授权编译自EE Times，版权所有，谢绝转载 第3页：各国法律都规定，车辆必须有驾驶人操控、且全时进行控制

相关阅读：
• 自动驾驶技术上路还有哪些问题？
• 自动驾驶还早，让我们先实现自动停车吧！
• 美国开放无人驾驶车测试，挂牌上路还需时日gcAesmc

{pagination} 而在被问到为何该机构认为自动驾驶车辆能从2020年的仅8,000台，在2035年大幅增加至9,540万台时，Alexander指出：“我们认为这就跟汽车在短短五年内快速采用自适应巡航技术一样，该市场在2000年时还非常小。” Strategy Analytics 所抱持的态度相对较保守，对此Riches解释：“以最新的预测数据来看，全球轻型车辆搭载各种自动巡航控制、车道偏离警报、盲点监测等功能的比例，在 2010年不到1%，在2015年估计可成长至5%，2020年为9%。”而他预测配备以上三种技术(也就是成为半自动或全自动驾驶车辆)的车辆比例，到 2025年为13%，2030年可增至18%。 而Riches也承认，这些预测数据：“可能是低估的，我们也预期市场的成长动力会随着相关技术成本降低，以及更多法规的颁布强制某种程度的驾驶人辅助系统之应用。”但无论是Navigant Research与Strategy Analytics都同意，在所有关于自动驾驶车辆的讨论中，产业界视而不见的一个问题，就是目前缺乏自动驾驶车辆的法规架构。 如Alexander所言，目前各国驾驶法规上的一大障碍，就是都规定所有车辆必须要有驾驶人操控、且全时进行控制：“在美国有一些州以及欧洲的某些国家，已经开始针对进行自动驾驶车辆发放测试驾照，允许该类车辆在受到控制的情况下于公共高速道路上行驶。” 这 类允许执照都需要新的立法；将此议题进一步复杂化，各家车厂都不会希望需要在自动驾驶车辆所造成的交通事故中负法律责任。但Riches 指出：“无论如何，汽车制造商还是得为所推出的产品背书；不过这些厂商在开始宣传并销售完全自动驾驶车辆之前，可能会需要去厘清并改变目前的产品责任法 规。” 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：Autonomous Cars: Breaking Down Market Forecasts，by Junko Yoshida

相关阅读：
• 自动驾驶技术上路还有哪些问题？
• 自动驾驶还早，让我们先实现自动停车吧！
• 美国开放无人驾驶车测试，挂牌上路还需时日gcAesmc

{pagination} Autonomous Cars: Breaking Down Market Forecasts Junko Yoshida MADISON, Wis. — How many "autonomous cars" do you think you'll see on the road worldwide in, say, 2030? Now, before getting caught up in the razzle-dazzle of Google self-driving cars, let's do the numbers we're fairly sure of. It turns out that opinions and forecasts among industry experts wildly vary -- ranging from an estimate of 20-30 million to 95 million autonomous cars around 2030 to 2035. This, however, depends on how autonomous people mean when they say "autonomous cars" and what sort of legal and commercial roadblocks they anticipate before the market reaches the nirvana of the self-driving car. On one end of the spectrum, Navigant Research of Boulder, Colo., offers perhaps the most optimistic prediction for self-driving cars. In a report called "Autonomous Vehicles," just released on Wednesday, Aug. 21, the market research firm pegged sales of autonomous vehicles to grow from fewer than 8,000 annually in 2020 to 95.4 million in 2035. Translated in a percentage of worldwide car sales, autonomous cars are expected to go up sharply from 0.01 per cent in 2020 to 75 per cent in 2035, David Alexander, senior research analyst with Navigant Research, told EE Times. In the predictions given above, Navigant Research defines "autonomous" as synonymous with the Google driverless car -- a vehicle that could essentially function as one's chauffeur. "I'm not talking about cars with some autonomous features built in, which could drive autonomously under limited circumstances," said Alexander. For example, the analyst is not putting cars like the upcoming 2014 Mercedes-Benz S-Class with freeway cruising features in the autonomous category, although some media outlets, such as CNN Money, called the Mercedes S-Class the "closest thing yet to a self-driving car." Boston-based Strategy Analytics, meanwhile, points out the big difference between a car with multiple advanced safety technologies and a truly autonomous vehicle. The market research firm expects autonomous cars that "are highly automated" (but not exactly self-driving) to have a market share of around 15 to 20 percent globally in 2025-2030. That translates into 20 million to 30 million autonomous cars. Ian Riches, director of Strategy Analytics' global automotive practice, defined those autonomous cars to be capable of "offering significant support to drivers in multiple different driving situations." But he told us the number of cars that are truly autonomous -- "that is, you can sit in your drive and program it to take you directly to the movie theatre, airport, or vacation destination" -- will be in the low single-figure percentages. Riches observed that fully autonomous cars "will likely only really emerge from 2025 onwards, and even then perhaps only initially offer full autonomous driving in certain situations, such as highway driving, or in areas with a certain degree of V2X support." Defining autonomy Because there is no "100 percent watertight definition" of autonomous cars today, as Riches pointed out, perhaps it's useful to describe different levels, according to five definitions put together by the US Department of Transportation's National Highway Traffic Safety Administration earlier this year. Level 0, according to the NHTSA, is a car with no automation. The driver is (we hope) in complete and sole control of the primary vehicle controls, such as brake, steering, throttle, and motive power, all the time. Level 1 indicates a car with one or more function-specific automation control functions -- including electronic stability control or pre-charged brakes. In other words, this allows the vehicle to automatically assist with braking to enable the driver to regain control of the vehicle or stop faster than possible by acting alone. Level 2, meanwhile, offers combined-function automation, offering at least two primary control functions designed to work in unison to relieve the driver. An example would be adaptive cruise control in combination with lane centering. Level 3 provides limited self-driving automation, enabling the driver to cede full control of all safety-critical functions under certain traffic or environmental conditions. The driver is expected to be available for occasional control, but with sufficiently comfortable transition time. The NHTSA sees the Google car as "an example of limited self-driving automation." Level 4 is defined by the NHTSA as "full self-driving automation." The vehicle is designed to perform all safety-critical driving functions and monitor roadway conditions for an entire trip. Such a design anticipates that the driver will provide destination or navigation input, but is not expected to be available for control at any time. Strategy Analytics' Riches noted, "I would start calling a vehicle 'semi-autonomous' at NHTSA's Level 2, progressing into level 3." Meanwhile, the term "fully autonomous" is reserved for Level 4 automation, he added. "Essentially, if the driver needs to be present and ready to take control if required, then the vehicle is only semi-autonomous." For any market research firm, predicting the future 20 years ahead is tricky business and inexact science. Riches described any forecast that extends to 2025 and beyond as "more in the realm of future-gazing than modeling." Navigant Research's Alexander, while acknowledging the challenge, noted, "Twenty years is a long way out, but we're making predictions based on the technologies that are already available," and information the firm has gathered on rollout plans by the industry. When asked about the huge jump in self-driving cars from 8,000 annually in 2020 to 95.4 million in 2035, Alexander said, "We see it no different from how cars with adaptive cruise control rapidly grew within a short of five years, from the early 2000s when the market was very small." Strategy Analytics, in contrast, takes a more conservative view. Riches explained: "Looking at the latest forecast data, the percentage of light vehicles worldwide fitted with all of autonomous cruise control, lane departure warning, and blind spot monitoring was less than 1 percent in 2010 and is projected at around 5 percent in 2015, rising to 9 percent in 2020." Riches further projected the percentage of vehicles with all three of these technologies (and thus the prospect of being partially/fully autonomous) at around 13 percent in 2025 and 18 percent in 2030. However, he acknowledged that these numbers could be "an underestimate, as we would expect growth to pick up as technologies get cheaper and more legislation is enacted mandating a certain level of driver support." The elephant in the room Both Navigant Research and Strategy Analytics agreed that the elephant in the room, in any discussion of self-driving cars, is the absence of a legal framework for autonomous cars. As Alexander described, the main barrier under the Geneva Convention on driving law, for all vehicles, is that drivers must be in charge of driving and be in control at all times. He explained, "Some US states and European countries have begun to issue licenses to companies to conduct testing of autonomous driving features on public highways under controlled conditions." But that permit requires new legislation. Complicating the matter further, automotive OEMs do not wish to be held liable for any accidents caused by self-driving cars. Riches noted: "Ultimately, it will be the vehicle manufacturer who has to stand behind their product. But they will probably need clarifications and likely [make] changes to current product liability laws before they can advertise and sell fully autonomous vehicles."