有些人不相信自动驾驶汽车将在可以预见的未来成为现实，对此，多位汽车厂商和大型供应商的高级技术总监给出了这样的回应：自动驾驶汽车的到来“就在眼前。”

在底特律附近举行的SAE 2016汽车电子大会（SAE 2016 Convergence）上，汽车产业供应商巨头德尔福（Delphi）首席技术官Jeff Owens给出了这样的回答，并赢得了几乎所有同行的赞同。福特公司（Ford）研发与先进工程副总裁Ken Washington也同意Owens的看法。

“证据就在我们周围，”Washington断言，“我坚信我们一定可以实现福特公司的2021年目标，即在2021年前推出一款符合SAE J3016标准定义的全自动驾驶汽车。”

Washington补充说，纵观汽车行业的历史，新技术的成熟速度，总是会让那些持怀疑态度的人大跌眼镜。众所周知，科技公司的创新速度快得惊人，再加上摩尔定律的作用，自动驾驶汽车时代的到来，可能会比人们想象的更快。

通用汽车(General Motors)副总裁、CTO兼投资主管Jon Lanckner表示，通用已经加快了自动驾驶技术的研发进程，并积极与更多相关的技术公司建立合作伙伴关系。他提到，通用汽车已经开始在美国加利福尼亚州的旧金山市和亚利桑那州的斯科茨代尔市进行自动驾驶汽车测试。

当然，还有一些反对的声音认为，自动驾驶汽车的真正到来，可能是下一代人才能看到的，甚至还会需要更长的时间。Lauckner在回应这一观点时幽默地打趣道，“我可以向你保证，我们绝不打算没完没了地一直测试下去。”

哈曼国际（Harman International）互联汽车执行副总裁Phillip Eyler也认为，自动驾驶汽车的到来将比大家想象的更快，但作为一家车载信息娱乐系统、驾驶员接口硬件/软件及其他车内电子产品供应商，哈曼也同时表示，自动驾驶汽车的出现是一回事，大规模普及又是另一回事。

Eyler表示，“我认为，自动驾驶汽车的普及还有很长的路要走。”

好的，自动驾驶就在眼前，但到底会有多“自动”呢？

目前，汽车和科技行业领袖探讨的焦点，主要围绕在到底何种自动驾驶级别最有利于自动驾驶汽车在最初阶段及日后发展中更为广泛的逐步普及。

大家的争论焦点主要集中在SAE自动驾驶分级中的 L3和L4两个级别。具体来说，L3级自动驾驶为“有条件的自动驾驶”，即可在特定环境下允许驾驶员“脱手”；L4级自动驾驶的应用范围则更加广泛，即可以在一些通常需要驾驶员干预的情况下，也能实现自动驾驶。

目前，许多汽车制造商、供应商和科技公司均委婉地表示，L3级自动驾驶在风险控制和收益回报方面的实现难度很大，而从L2级驾驶员协助系统直接过渡至L4级全自动系统已经可以实现。

福特的Washington在大会上表示，实现高度自动化驾驶“需要采用自上而下，而非自下而上的措施。”他还补充道，福特已经将公司向L2级驾驶员辅助系统的投资增加了2倍。“福特准备直接从L2跳到L4级全自动驾驶，”他说，“L3级对我们来说没有什么经济效益。”

哈曼的Eyler也持相同观点，他说，“L4级全自动驾驶使用了一套完全不同的技术，投资规模也完全不在一个级别。”不过，Eyler也相信，还是会有一部分厂商可能不会跳过L3级的过渡。

目前，从多个层面来看，SAE L3级自动驾驶相关技术的研发已经有点得不偿失。厂商第一个要考虑的就是同时开发L3和L4级技术的成本压力。从技术研发的经济性来说，“L3级自动驾驶一直是个有争议的话题，”德尔福的Owens表示，“你必须直接跳过L3级的半自动驾驶级别。”

另外，优步（Uber）、Lyft等驾乘分享公司也希望利用自动驾驶技术节省雇佣人类司机的成本，因此也为直接跳级实现L4级全自动驾驶的热潮添了一把火。考虑到无人驾驶叫车服务的巨大市场潜力，汽车行业根本没有理由在困难重重且成本高昂的L3级半自动驾驶技术研发上继续恋战。

直接买还是自己造？

在SAE 2016汽车电子大会与展览上，专家们还探讨了汽车生产商可能采取的自动驾驶战略，比如内部研发、寻求技术公司合作，或者直接收购一些自动驾驶技术公司。汽车制造商开始大举收购技术公司，已经有一段时间了。

通用汽车的Lauckner表示，“我们可以花大量时间延续过去的模式，也就是进行内部研发。”但他同时补充，未来公司将更多采用“直接收购特定技术公司的新模式”。

福特的Washington表示，因为拥有一百多年积累下来的汽车研发经验，汽车制造商的整合能力很强。他说，“而我们并不拿手的技术研发，也可以通过建立合作伙伴关系或收并购解决。”

德尔福的Owens还提出了另一种大型供应商和较小汽车制造商希望尝试的模式，即由供应商充当一站式自动驾驶系统提供商。他指出，德尔福已在2016年夏季宣布将与以色列机器视觉专家Mobileye建立合作伙伴转系，联合开发完整自动驾驶系统。

For those who believe it unlikely autonomously-piloted vehicles will be coming in the foreseeable future, high-level technology executives from automakers and major suppliers have this response: autonomous vehicles are “going to be here very shortly.”

That’s the conclusion of Jeff Owens, Chief Technology Officer for automotive mega-supplier Delphi, a position supported almost universally by his fellow panelists on the opening day of the SAE 2016 Convergence conference near Detroit. Ken Washington, Vice-President, Research and Advanced Engineering at Ford, agreed with Owens.

“The evidence is all around us," Washington asserted. "I am very optimistic that [Ford’s] target of 2021 (for launching a fully-autonomous vehicle as defined by the SAE’s J3016 standard) is going to be achievable.”

Washington added that throughout automotive history, the speed at which technology matures never fails to surprise doubters. Blend in the famously furious pace at which technology companies innovate, along with Moore's Law, and the foundation seems prepared for a quicker-than-predicted adoption.

General Motors has accelerated its autonomous-technology development and partnerships with autonomous-related tech companies, said Jon Lauckner, GM Vice President, CTO and head of GM Ventures. He noted that GM has autonomous-vehicle testing underway on public roads in San Francisco, CA and Scottsdale, AZ.

“I can assure you our plan is not to test for the next 30 years,” Lauckner quipped in response to naysayers’ position that it will be a generation or more before autonomous vehicles are a fixture on U.S. roads.

Phillip Eyler, Executive Vice President, Connected Car at Harman International, generally agreed that autonomous vehicles are coming sooner than later, but the position of Harman—a supplier of in-car infotainment systems, driver-interface hardware and software and other cabin electronics—is that the date of introduction may be one thing, but wide adoption of autonomy will come on a more-protracted timeframe.

“I believe there will be a long transition to a high population of autonomous cars,” Eyler said.

Okay, autonomous—but at what level?

One of the automotive and tech industries’ chief talking points currently centers around which SAE level of autonomy is going to be most appropriate for initial—and eventual widespread—deployment.

The dialogue centers on the gap between SAE Level 3 automation—generally described as “conditional automation” that can necessitate a “hand-off” from automation to human driving in certain situations—and Level 4 autonomy which can pilot the vehicle even in a situation in which the system requests human intervention.

Many automakers, suppliers and tech companies now suggest that Level 3 autonomy is too difficult to engineer in relation to the risk (and payback), instead suggesting a transition from the Level 2 driver-assistance systems already available directly to Level 4.

High-level driving automation “needs to be approached top-down instead of bottom-up,” said Ford’s Washington. In the interim, he added, Ford is tripling its investment in Level 2 driver-assist technology. “Ford is skipping Level 3 at the moment,” he said. “The economics don’t make sense to us.”

Harman’s Eyler agreed.“Level 4 is a totally different set of technologies and investments,” he stated. However there will be a Level 3 transition at some OEMs, Eyler believes.

The cost of developing technology capable of accommodating SAE Level 3 automation is beginning to appear unproductive on several levels. First is the direct cost of the divergent Level 3 and Level 4 technologies. Economics of such development “are going to make Level 3 almost a moot point,” contended Delphi’s Owens. “You’re going to go right by Level 3.”

But the impetus for skipping directly to Level 4 also is driven by the desire of ride-sharing companies such as Uber, Lyft and others to eliminate the expense of human drivers. Given the potential enormous market for driverless ride-hailing vehicles, it seems all the more logical that development will not linger on the difficult and expensive-to-engineer conditional-automation aspects of Level 3.

Buy it or make it?

Panelists at SAE Convergence 2016 also discussed the strategy by which automakers determine whether to develop autonomous-related technologies in-house, seek partnerships with tech companies or, in some cases, to buy some companies outright. Automakers have for some time been underway with a spree of tech-company acquisitions.

GM’s Lauckner said “We can spend a lot of time working the ‘old model,’ which is build it ourselves.” But he added, buying companies with specific technical expertise “is a model we’re going to need to embrace.”

Thanks to a century of vehicle-development experience, automakers are ideally placed to be high-level integrators said Ford’s Washington. “It’s what we’re good at,” he said, while development of technology that is “not in our sweet spot of capability” will be partnered or potentially bought outright.

Delphi’s Owen’s pointed to another model that large suppliers and smaller automakers hope to exploit: suppliers as providers of turnkey autonomous-driving systems. He points to Delphi’s partnership, announced in summer 2016, with Israeli machine-vision expert Mobileye to develop entire autonomous-driving systems.

Author: Bill Visnic
Source: SAE Automotive Engineering Magazine