Elon Musk再次對無人駕駛出租車夸下???。
十多年來,涉足自動(dòng)駕駛汽車的各公司均遇到了遠(yuǎn)超最初預(yù)期的難題。因此,該領(lǐng)域遍布著半途而廢的項(xiàng)目。
盡管競爭對手接連折戟,特斯拉本身也屢次未能如期推出自動(dòng)駕駛出租車,但其決心似乎并未動(dòng)搖。在加州伯班克華納兄弟影城舉辦的“We, Robot”發(fā)布會現(xiàn)場,Elon Musk發(fā)布了兩款專用無人駕駛出租車,并公布了其最新上市時(shí)間。
第一款車型為CyberCab兩座無人駕駛出租車。這款小型自動(dòng)駕駛汽車融合了Cybertruck電動(dòng)皮卡和本田CR-Z的設(shè)計(jì),計(jì)劃于2026年投產(chǎn)。然而,Musk也承認(rèn)其設(shè)定的投產(chǎn)時(shí)間往往過于樂觀。目前為止,特斯拉未能如期投產(chǎn)的項(xiàng)目數(shù)量遠(yuǎn)超其如期投產(chǎn)的項(xiàng)目。
SAE有幸獲得了試乘CyberCab的機(jī)會。其小巧的體型和鷗翼式車門在特斯拉于華納兄弟影城搭建的模擬城市場景中顯得十分和諧。車內(nèi)空間對兩名乘客來說足夠?qū)挸?,座椅也相?dāng)舒適。片場環(huán)境使車輛能夠按照既定路線行駛,也沒有真實(shí)城市的喧囂與危險(xiǎn),因此車輛的方向盤和踏板理論上并不存在多大風(fēng)險(xiǎn)。但是,這樣的設(shè)計(jì)仍然令一些有意試乘的觀眾望而卻步。
大屏幕上播放著科幻電影片段與80年代的經(jīng)典配樂,為CyberCab營造了一個(gè)展示技術(shù)潛力的舞臺。然而,截至目前,特斯拉尚未正式推出任何一款自動(dòng)駕駛汽車。
CyberCab完全依賴攝像頭視覺技術(shù),這是特斯拉備受爭議的一項(xiàng)決策。大多數(shù)致力于自動(dòng)駕駛技術(shù)研發(fā)的公司都會結(jié)合使用攝像頭、雷達(dá)和激光雷達(dá)進(jìn)行導(dǎo)航。然而,Musk在向粉絲宣布CyberCab的售價(jià)將低于3萬美元的同時(shí),還強(qiáng)調(diào),“這款車完全依賴人工智能視覺技術(shù),因此不需要昂貴的設(shè)備。”
喬治梅森大學(xué)梅森自動(dòng)駕駛與機(jī)器人中心主任Missy Cummings教授在接受SAE采訪時(shí)表示:“我認(rèn)為僅憑計(jì)算機(jī)視覺技術(shù),不可能實(shí)現(xiàn)穩(wěn)定可靠的自動(dòng)駕駛,甚至駕駛輔助功能。”
Cummings表示:“我認(rèn)為本次發(fā)布會毫無新意。特斯拉要想在其描述的網(wǎng)絡(luò)環(huán)境中實(shí)現(xiàn)真正的自動(dòng)駕駛,至少還需要十年的時(shí)間。不過我很喜歡特斯拉的20座廂式車Robovan,我認(rèn)為它應(yīng)該重點(diǎn)研究該項(xiàng)目。從經(jīng)濟(jì)效益的角度而言,這種車型的研發(fā)更具意義,而且在短期內(nèi)可能更容易成功(前提是特斯拉不再堅(jiān)持完全依賴視覺技術(shù))。”
特斯拉突然轉(zhuǎn)向研發(fā)20座無人駕駛廂式車的舉動(dòng)令人十分意外,畢竟Musk曾對公共交通表達(dá)出不屑的態(tài)度。這款廂式車旨在緩解通勤或前往機(jī)場途中的交通擁堵。
不過,在一場SAE參與的視頻會議中,密歇根大學(xué)Mcity常務(wù)董事Greg McGuire指出,特斯拉發(fā)布會上并未提及任何關(guān)于無障礙通行(特別是針對乘坐輪椅的乘客)的內(nèi)容。McGuire對CyberCab的創(chuàng)意大為贊賞。他表示:“我認(rèn)為兩座的設(shè)計(jì)頗具意義,因?yàn)閱螐钠嚺c人體的動(dòng)能比來看,體積更小、重量更輕的汽車必定有助于打造一個(gè)更安全的城市。汽車的重量越輕或速度越慢,對自行車騎行者和行人來說安全性就會越高。”
但McGuire也表達(dá)了與Cummings相同的擔(dān)憂,認(rèn)為特斯拉的自動(dòng)駕駛技術(shù)恐怕到2027年也無法準(zhǔn)備就緒。McGuire表示:“據(jù)我了解,特斯拉全自動(dòng)駕駛系統(tǒng)的性能目前還未達(dá)到L4級自動(dòng)駕駛的一般標(biāo)準(zhǔn)。到2027年,他們真能達(dá)到L4級水平嗎?在Mcity的模擬測試環(huán)境中,我們發(fā)現(xiàn)特斯拉的技術(shù)仍存在一些關(guān)鍵的技術(shù)障礙。”
多年來,特斯拉方面不斷承諾自動(dòng)駕駛汽車和無人駕駛出租車即將面世。2016年,首席執(zhí)行官Elon Musk在一次財(cái)報(bào)視頻會議和一篇關(guān)于特斯拉發(fā)展規(guī)劃的博客文章中均提到了這一承諾(現(xiàn)已刪除)。2019年,他宣布這些汽車將于當(dāng)年年底上路。2020年,Musk再次承諾無人駕駛出租車將于當(dāng)年年底準(zhǔn)備就緒。
公眾質(zhì)疑是可以理解的。太平洋汽車公司(AutoPacific)的產(chǎn)品和消費(fèi)者分析經(jīng)理Robby DeGraff在發(fā)給SAE的電子郵件中提到:“回顧他對完全自動(dòng)駕駛(FSD)、Roadster電動(dòng)跑車、Cybertruck皮卡做出的種種承諾,公眾的疑慮也在情理之中。如果安全性能夠得到保障,我還是非??春眠@項(xiàng)技術(shù)的前景的。但我認(rèn)為CyberCab的價(jià)格不可能低于3萬美元,而且也不太可能在Musk設(shè)定的如此緊迫的時(shí)間投產(chǎn)。因此,關(guān)于價(jià)格和投產(chǎn)時(shí)間的承諾,我們大可不必當(dāng)真。”
雖然特斯拉可能對無人駕駛出租車持樂觀態(tài)度,但公眾可能尚未準(zhǔn)備好接受這種交通工具。DeGraff表示:“在CyberCab正式上路之前,特斯拉仍需完成大量工作,比如滿足受監(jiān)控全自動(dòng)駕駛汽車相關(guān)的法規(guī)要求,更重要的是贏得公眾信任并令其安心。我們的研究顯示,消費(fèi)者普遍對更高級的自動(dòng)駕駛技術(shù)持謹(jǐn)慎態(tài)度,尤其是他們自己車上使用的技術(shù)。因此,可以說公眾對無人駕駛出租車的需求并不高。”
結(jié)束CyberCab試駕后,隨著車門緩緩打開,我們也下了車。盡管特斯拉的目標(biāo)是展示完全無人駕駛,但這些車輛實(shí)際上還是會受到遠(yuǎn)程監(jiān)控。當(dāng)遇到難以決策的情況時(shí),人類駕駛員必須接管車輛,以避免類似于Cruise自動(dòng)駕駛出租車在舊金山造成的堵車事故變成常態(tài)。漫步于影城中時(shí),我們注意到,在矗立著巨大屏幕的特斯拉燈光秀舞臺后方的一棟建筑中,其員工正在密切監(jiān)控?cái)?shù)面墻上的監(jiān)視器。
監(jiān)視器屏幕上顯示了CyberCab和示范中的其它特斯拉自動(dòng)駕駛汽車的攝像頭捕捉的畫面。工作人員正在實(shí)時(shí)監(jiān)控這些車輛。這就是自動(dòng)駕駛實(shí)際面臨的問題之一 —— 必須受到遠(yuǎn)程監(jiān)控并配合臨時(shí)的人為干預(yù),才能確保安全駕駛。
發(fā)布會現(xiàn)場的大屏幕上播放著特斯拉設(shè)想的自動(dòng)駕駛世界的宣傳片,其中展示了車主可以在CyberCab的行駛過程中觀看電影的情節(jié),以及車主在接送鄰居時(shí)賺取一些額外報(bào)酬的情節(jié)。然而,現(xiàn)實(shí)情況往往更為復(fù)雜,現(xiàn)實(shí)的交通狀況也比模擬環(huán)境更為混亂。特斯拉必須制造出能夠在現(xiàn)實(shí)世界的各種復(fù)雜環(huán)境中安全行駛的汽車。
只是,Musk再次設(shè)定了一個(gè)許多人都認(rèn)為不切實(shí)際的交付時(shí)間。
For over a decade, the autonomous vehicle problem has been far more difficult than originally expected for every company involved. The landscape is littered with abandoned attempts.
Tesla doesn't seem to be fazed by the failure of competitors and even its own many missed deadlines to bring autonomous taxis to the market. At its We, Robot event at the Warner Bros. studio in Burbank, CA, Musk appeared onstage to unveil a pair of dedicated robotaxis and a new timetable for the availability of autonomous vehicles.
The first vehicle is the two-seater Cybercab. A cross between the Cybertruck and the Honda CR-Z, the small autonomous vehicle is set to go into production in 2026. Although, Musk did admit a tendency to be optimistic with timelines. At this point in its history, more public Tesla deadlines have been missed than kept.
SAE Media had a chance to ride in the Cybercab. Its small stature and gullwing doors did feel in place in the faux-city Tesla put together on the Warner Bros. lot. The interior was roomy for two people, and the seats were comfortable. The lack of a steering wheel or pedals, while not alarming in the operating domain of a movie set with prescribed routes without the hustle and bustle (and dangers) of a real city, could make some of the public pause before riding.
A large screen offered up bits of sci-fi movies, and a soundtrack straight out of the ‘80s blared in our ears. It was a demo of the potential of the vehicle. As of right now, Tesla does not have a self-driving vehicle on the market.
One of the more controversial decisions by Tesla is to rely entirely on camera vision. Most other companies working on automated driving are using cameras in conjunction with radar and lidar to navigate the world. "I want to emphasize that the solution that we have is AI vision. So there's no expensive equipment needed," Musk told the crowd of fans as he announced that the CyberCab would cost under $30,000.
"I do not think self-driving (or even driving assist) can be reliably accomplished with computer vision." Missy Cummings, professor and director of Mason Autonomy and Robotics Center at George Mason University, told SAE.
"I saw nothing new in the reveal,” Cummings said. “Tesla is at least 10 years away from achieving real self-driving in the network they describe. But I did love their 20-person shuttle and wish they would focus on this. The economies of scale make more sense and they likely could be more successful in the short term (but only if they stop insisting on vision only).”
And a Robovan?
The pivot to a 20-person Robovan was a surprise. Musk has noted his disdain for public transportation in the past. The van would reduce congestion during commutes or trips to the airport. However, during a call with SAE Media, Greg McGuire, managing director of Mcity at the University of Michigan, noted that there was zero mention of accessibility (particularly for wheelchair-bound riders) at the event.
McGuire did like the idea of the CyberCab. "I think the two-seater idea is actually pretty interesting because a smaller form factor vehicle, lighter weight, will definitely contribute to a safer city in terms of just the sheer kinetic energy of a vehicle versus a human. Either bicycle or pedestrian, you know, the less that vehicle weighs, or the slower it's going, the better."
McGuire also echoed Cummings’ concern that the Tesla technology would not be ready by 2027. "The performance of Tesla's full self-driving stack is not – from what I've seen – ready for general Level 4 operation. Will they be there by 2027? At Mcity, we still think there's a couple of key scientific barriers," McGuire said.
Tesla, for its part, has been promising that autonomous vehicles and robotaxis are just around the corner for years. In 2016, CEO Elon Musk mentioned it during an earnings call and in his master plan blog post, which has since been deleted. In 2019, Musk announced that the vehicles would be on the road by the end of that year. In 2020, Musk again said that the robotaxis would be ready by the end of that year.
Being skeptical of Musk's claims is understandable. "Recall all the past promises about FSD, the Roadster, the Cybertruck, etc. While I’m excited about the prospect of the technology –should it actually prove to be safe – I don’t think it’s realistically going to be available at a sub-$30k cost, nor will the Cybercab enter production within the ambitious timeline Musk suggested. You can pretty much toss both of those comments right out the window," Robby DeGraff, manager of product and consumer insights at AutoPacific, told SAE Media in an email.
While Tesla might be bullish on robotaxis, the general public might not be ready, "There’s much work to be done before these things hit the road, regulations to be met, and, equally important, public trust and assurance need to be gained for supervised FSD. Generally speaking, our research shows that consumers remain very hesitant about more advanced forms of autonomy in vehicles, especially ones of their own. There’s not much demand for it," DeGraff said.
Behind the CyberCab
At the end of the ride in the CyberCab, the door swung open and we exited the vehicle. While Tesla wants to do away with individual drivers, these vehicles are typically monitored remotely, and when they get confused, a human needs to interact so that situations like those that have happened in San Francisco with Cruise vehicles blocking the roadway don't become the norm. While walking on the studio lot, in the building behind where Musk and Tesla had set up a stage with enormous screens and a light show, we noticed walls of monitors being watched.
On the screens was the view of the world from the cameras of the CyberCabs and other Tesla models participating in the autonomous demo. Humans were keeping tabs on the vehicles. This is one of the realities of autonomous vehicles. Remote monitoring of the vehicles themselves with occasional human intervention to keep the vehicles on track.
A movie studio made for a great backdrop of an autonomous world where traffic just means more time to watch a film and every car is making someone some extra cash while ferrying their neighbors around. The reality is much tougher, and the world is far more chaotic. Tesla needs to build vehicles that can navigate the real world with all its insanity and, once again, Musk set himself up with a timeline that many don't believe is possible.