以谷歌无人车的视角来看无人驾驶 The View from the Front Seat of the Google Self-Driving Car

该文的英文版由现在谷歌无人车项目负责人Chris Urmson撰写,并发布在BackChannel上。崔迪潇完成该文的中文翻译。English version of the post is by Chris Urmson, director of Google’s Self-driving car program, and it is originally published on BackChannel. The Chinese version is translated by Dixiao Cui.

我们从170万英里的驾驶中学到了很多,既有对系统的认识也包含对人的驾驶行为的认识。After 1.7 million miles we’ve learned a lot — not just about our system but how humans drive, too.

每年美国约有33,000人死于道路交通事故。这也是为何我们一直强调和展望无人驾驶汽车在减少交通事故方面的前景。当我们不断接近一键实现从A点载客到达B点的全自动无人驾驶愿景时,我们也在考虑应当如何评价所取得的进步,以及这些进步对道路安全带来的影响。为了评价我们车辆的安全性能,我们需要分析的一个重要指标是车辆在典型城郊道路上面对事故时的性能基准。由于在官方数据中缺少很多事故分析数据,我们需要自己分析出无人驾驶汽车被其他车辆碰撞的可能性。即便软件系统可以检测到一些棘手的交通场景,并能比一个反应敏捷的驾驶员更早更快的做出对应操作,但这仍不能克服车辆本身的速度和预警距离带来的限制;某些时候仅仅是因为无人车在等待交通灯就会被其他车撞上。这点对于有无人驾驶汽车在路上行驶的地区来说尤其重要;尽管我们希望无人驾驶汽车能够避免所有事故,但仍有一些事故不可避免。About 33,000 people die on America’s roads every year. That’s why so much of the enthusiasm for self-driving cars has focused on their potential to reduce accident rates. As we continue to work toward our vision of fully self-driving vehicles that can take anyone from point A to point B at the push of a button, we’re thinking a lot about how to measure our progress and our impact on road safety. One of the most important things we need to understand in order to judge our cars’ safety performance is “baseline” accident activity on typical suburban streets. Quite simply, because many incidents never make it into official statistics, we need to find out how often we can expect to get hit by other drivers. Even when our software and sensors can detect a sticky situation and take action earlier and faster than an alert human driver, sometimes we won’t be able to overcome the realities of speed and distance; sometimes we’ll get hit just waiting for a light to change. And that’s important context for communities with self-driving cars on their streets; although we wish we could avoid all accidents, some will be unavoidable.

对于无人驾驶汽车而言,绝大部分可能遇到的日常交通事故都是没有人受伤的轻微刮蹭,但由于通常这些事故并没有通报警察,对这一类事故的发生原因我们实际上知之甚少。NHTSA的数据表明,55%的碰撞属于上述这一类事故。如果没有每天数英里的驾驶,是很难理解在实际的道路驾驶中究竟发生了什么。而这也正是我们目前在努力解决的事情。我们车队一共有超过20辆无人驾驶汽车,以及一组安全可靠的驾驶员,目前人工驾驶和自动驾驶已经总共累积行驶了170万英里。其中约100万英里为无人驾驶,目前平均每周的无人驾驶距离约为10000英里,且绝大部分路程在城区环境中。The most common accidents our cars are likely to experience in typical day to day street driving — light damage, no injuries — aren’t well understood because they’re not reported to police. Yet according to National Highway Traffic Safety Administration (NHTSA) data, these incidents account for 55% of all crashesIt’s hard to know what’s really going on out on the streets unless you’re doing miles and miles of driving every day. And that’s exactly what we’ve been doing with our fleet of 20+ self-driving vehicles and team of safety drivers, who’ve driven 1.7 million miles (manually and autonomously combined). The cars have self-driven nearly a million of those miles, and we’re now averaging around 10,000 self-driven miles a week (a bit less than a typical American driver logs in a year), mostly on city streets.

为了帮助大家都更加安全地驾驶,我们在此分享几个典型的、有人为操作失误的道路交通场景。其中大多数相信你们已经见过,尤其是如果你知道94%的交通事故是由驾驶员操作失误导致的。In the spirit of helping all of us be safer drivers, we wanted to share a few patterns we’ve seen. A lot of this won’t be a surprise, especially if you already know that driver error causes 94% of crashes.

如果你在路上驾驶的时间足够多,无论你是人工驾驶还是身处无人车,总难免会发生事故。在过去6年间,我们在170万英里、有安全驾驶员监督的人工和自动驾驶路程中一共发生了11次小的事故(轻微损坏,无人受伤),并且没有一次是无人车的原因导致的事故。If you spend enough time on the road, accidents will happen whether you’re in a car or a self-driving car. Over the 6 years since we started the project, we’ve been involved in 11 minor accidents (light damage, no injuries) during those 1.7 million miles of autonomous and manual driving with our safety drivers behind the wheel, and not once was the self-driving car the cause of the accident.

追尾是美国最普遍的交通事故,通常前方车辆司机基本上对避免事故无能为力;我们共被追尾7次,大多数发生在路口等待交通灯时,也在高速路发生过。我们也被侧面碰撞了几次。如你所预料,城区道路每英里的事故数量比高速路多;我们共有8次事故,是在城区道路上刚行驶没多久就被其他车辆撞了。这些疯狂的经历对我们的项目益处颇多。即便并不是我们的无人驾驶汽车的错,我们仍然启用了一套细致的事故分析流程,并尽可能从每次事故中总结经验,。Rear-end crashes are the most frequent accidents in America, and often there’s little the driver in front can do to avoid getting hit; we’ve been hit from behind seven times, mainly at traffic lights but also on the freeway. We’ve also been side-swiped a couple of times and hit by a car rolling through a stop sign. And as you might expect, we see more accidents per mile driven on city streets than on freeways; we were hit 8 times in many fewer miles of city driving. All the crazy experiences we’ve had on the road have been really valuable for our project. We have a detailed review process and try to learn something from each incident, even if it hasn’t been our fault.

我们不仅在认真分析城郊道路上发生轻微事故的可能性,同时也分析了一些可能导致严重碰撞的典型驾驶员行为模式 (偏离车道,闯红灯等)。这些行为数据在官方资料里很难查到,但却对道路安全造成了威胁。Not only are we developing a good understanding of minor accident rates on suburban streets, we’ve also identified patterns of driver behavior (lane-drifting, red-light running) that are leading indicators of significant collisions. Those behaviors don’t ever show up in official statistics, but they create dangerous situations for everyone around them.

多数驾驶员并不关注道路状况。在白天的任一时段中,全美约有660,000的驾驶员在驾驶中只顾翻看电子设备而不盯着车前道路。我们的安全驾驶员遇到过车辆在不同车道间来回换道,驾驶员在车内看书,甚至有人吹小号。无人驾驶汽车在道路安全性能上远胜于人类,它具备360度感知覆盖、以及全天候全方位的检测;目前我们最新的传感器能够跟踪两个足球场远的车辆、自行车和行人。Lots of people aren’t paying attention to the road. In any given daylight moment in America, there are 660,000 people behind the wheel who are checking their devices instead of watching the road. Our safety drivers routinely see people weaving in and out of their lanes; we’ve spotted people reading books, and even one playing a trumpet. A self-driving car has people beat on this dimension of road safety. With 360 degree visibility and 100% attention out in all directions at all times; our newest sensors can keep track of other vehicles, cyclists, and pedestrians out to a distance of nearly two football fields.

路口是危险地带。过去几年间,全美道路上21%的死亡和约50%的严重事故发生在路口。行人和其他驾驶员往往是事故受害者,而非乱闯红灯的违章驾驶员。在交通灯变绿后,经常会有某些驾驶员在这个时间及不耐烦或注意力不集中地驶过路口,因此我们设定无人驾驶汽车在检测到交通灯变绿后仍然需要短暂等待一段时间后,才能进入路口。Intersections can be scary places. Over the last several years, 21% of the fatalities and about 50% of the serious injuries on U.S. roads have involved intersections.  And the injuries are usually to pedestrians and other drivers, not the driver running the red lightThis is why we’ve programmed our cars to pause briefly after a light turns green before proceeding into the intersection — that’s often when someone will barrel impatiently or distractedly through the intersection.

如上图所示,一个骑自行车的人(淡蓝色矩形)没有及时进入路口,而在路口停留的一辆车(正在进入路口的紫色矩形框)的驾驶员并没有看到该自行车,并在交通灯变绿后开始发动车辆左转,差点就将该自行车撞倒。我们的无人驾驶汽车成功地预测了自行车驾驶员的行为(红色路线),直到该自行车驾驶员安全通过路口后才启动。In this case, a cyclist (the light blue box) got a late start across the intersection and narrowly avoided getting hit by a car making a left turn (the purple box entering the intersection) who didn’t see him and had started to move when the light turned green. Our car predicted the cyclist’s behavior (the red path) and did not start moving until the cyclist was safely across the intersection.

转弯可能导致麻烦。我们经常遇到驾驶员转入错误道路行驶,尤其是晚上驾驶员常常难以保证行驶在安全岛的正确一侧。Turns can be trouble. We see people turning onto, and then driving on, the wrong side of the road a lot — particularly at night, it’s common for people to overshoot or undershoot the median.

上图中所示的两辆车辆(绿色路线左侧的两个紫色矩形框,如上图的左下图所示)从安全岛的错误一侧朝我们驶来;这发生在山景城最繁忙的夜间道路上。In this image you can see not one, but two cars (the two purple boxes on the left of the green path are the cars you can see in the photo) coming toward us on the wrong side of the median; this happened at night on one of Mountain View’s busiest boulevards.

此外,当驾驶员意识到需要立刻转弯时,他们常常会做出一些2B的事情。Other times, drivers do very silly things when they realize they’re about to miss their turn.

带有感叹号的紫色矩形框车辆决定从我们所在车道的左侧右转进入路口,严重阻碍了我们车的路径。而这些绿色矩形框,我们称之为栅栏,表明我们的无人驾驶汽车将减速来躲避该车。A car (the purple box touching the green rectangles with an exclamation mark over it) decided to make a right turn from the lane to our left, cutting sharply across our path. The green rectangles, which we call a “fence,” indicate our car is going to slow down to avoid the car making this crazy turn.

其余时候,有些车辆在驾驶时似乎完全没有注意到无人驾驶汽车也在路上。下图所示,最左侧车道上有一辆被被红色栅栏穿过的紫色矩形框所示车辆强行转向,并阻挡了我们的无人驾驶汽车。红色栅栏表明我们的无人驾驶汽车自动停下,躲避该车。And other times, cars seem to behave as if we’re not there. In the image below, a car in the leftmost turn lane (the purple box with a red fence through it) took the turn wide and cut off our car. In this case, the red fence indicates our car is stopping and avoiding the other vehicle.

这些经历(以及其他更多的类似经历)不断加深了我们对每天所面临的挑战的认识。我们将继续千里驰骋,以期待更好的理解和分析在日常驾驶中常见的烦人的交通事故,并会继续研发无人驾驶汽车来减轻我们的驾驶负担。These experiences (and countless others) have only reinforced for us the challenges we all face on our roads today. We’ll continue to drive thousands of miles so we can all better understand the all too common incidents that cause many of us to dislike day to day driving — and we’ll continue to work hard on developing a self-driving car that can shoulder this burden for us.

关于dixiaocui

无人驾驶汽车研究员 半吊子摇滚混子乐手 Researcher on Intelligent Vehicle Guitarist/Singer/Rocker
此条目发表在Autonomous Driving分类目录,贴了, , , , , 标签。将固定链接加入收藏夹。

发表评论

电子邮件地址不会被公开。

此站点使用Akismet来减少垃圾评论。了解我们如何处理您的评论数据