TED2015
Chris Urmson: How a driverless car sees the road
克里斯·厄姆森: 无人驾驶汽车是如何看清路况的
Filmed:
Readability: 3.3
2,536,355 views
统计上来说，汽车上最不可靠的部分是...驾驶员。克里斯·厄姆森带领的谷歌无人汽车驾驶项目，是几个想把人类从驾驶座上移除的项目之一。他谈论了目前项目的进展，并分享了有趣的视频，为我们展示了这类汽车如何分析路况以及如何自动做出下一步的抉择。
Double-click the English transcript below to play the video.
00:12
So in 1885, Karl Benz
invented the automobile.
invented the automobile.
0
528
3949
在1885年，卡尔本茨发明了汽车。
00:16
Later that year, he took it out
for the first public test drive,
for the first public test drive,
1
4707
3762
那年年末，他把车开出去
进行第一次户外测试,
进行第一次户外测试,
00:20
and -- true story --
crashed into a wall.
crashed into a wall.
2
8469
3375
然后——真实是——他撞墙了。
00:24
For the last 130 years,
3
12184
2043
在过去的130年里，
00:26
we've been working around that least
reliable part of the car, the driver.
reliable part of the car, the driver.
4
14227
4319
我们一直在为汽车的最薄弱环节，
驾驶员，做着努力。
驾驶员，做着努力。
00:30
We've made the car stronger.
5
18546
1354
我们把车做得更坚固,
00:32
We've added seat belts,
we've added air bags,
we've added air bags,
6
20200
2548
我们增加了安全带，增加了安全气囊。
00:34
and in the last decade, we've actually
started trying to make the car smarter
started trying to make the car smarter
7
22748
3971
在过去十年里，
我们开始让车变得更智能,
我们开始让车变得更智能,
00:38
to fix that bug, the driver.
8
26719
2938
是为弥补驾驶员方面的缺陷。
00:41
Now, today I'm going to talk to you
a little bit about the difference
a little bit about the difference
9
29657
3261
那么，今天我将给大家讲讲，
00:44
between patching around the problem
with driver assistance systems
with driver assistance systems
10
32918
3808
用驾驶员辅助系统解决问题，
00:48
and actually having fully
self-driving cars
self-driving cars
11
36726
2564
和真正使用全面自动驾驶汽车的区别，
00:51
and what they can do for the world.
12
39290
1880
以及它们对世界的贡献。
00:53
I'm also going to talk to you
a little bit about our car
a little bit about our car
13
41170
2995
我也会介绍一下我们的车，
00:56
and allow you to see how it sees the world
and how it reacts and what it does,
and how it reacts and what it does,
14
44165
3999
让你们了解它是如何观察这个世界的，
以及如何应对各种情况，
以及如何应对各种情况，
01:00
but first I'm going to talk
a little bit about the problem.
a little bit about the problem.
15
48164
3187
但首先我先说说存在的问题。
01:03
And it's a big problem:
16
51651
1648
这是个很严重的问题，
每年全世界都有120万人
因交通事故丧命。
因交通事故丧命。
01:05
1.2 million people are killed
on the world's roads every year.
on the world's roads every year.
17
53299
3089
01:08
In America alone, 33,000 people
are killed each year.
are killed each year.
18
56388
3784
仅仅在美国，
每年就有3万3千人死于车祸。
每年就有3万3千人死于车祸。
01:12
To put that in perspective,
19
60172
2028
换个方式说,
01:14
that's the same as a 737
falling out of the sky every working day.
falling out of the sky every working day.
20
62200
4797
等同于每天都有一架737飞机失事。
01:19
It's kind of unbelievable.
21
67342
1786
有点不可思议。
01:21
Cars are sold to us like this,
22
69548
2298
汽车卖给我们后应该是这样一番景象，
01:23
but really, this is what driving's like.
23
71846
2717
但事实上，驾驶过程通常是这样。
01:26
Right? It's not sunny, it's rainy,
24
74563
2159
对吧？这不是晴天，是雨天，
01:28
and you want to do anything
other than drive.
other than drive.
25
76722
2488
而且除了开车，
你还想做点别的事情。
你还想做点别的事情。
01:31
And the reason why is this:
26
79210
1622
原因就是：
01:32
Traffic is getting worse.
27
80832
1858
交通状况变得越来越糟。
01:34
In America, between 1990 and 2010,
28
82690
3506
在美国，从1990年到2010年,
01:38
the vehicle miles traveled
increased by 38 percent.
increased by 38 percent.
29
86196
3504
交通工具的里程数增加了38%。
01:42
We grew by six percent of roads,
30
90213
2749
而我们只增修了6%的路,
01:44
so it's not in your brains.
31
92962
1602
所以不单单是你的感觉如此。
01:46
Traffic really is substantially worse
than it was not very long ago.
than it was not very long ago.
32
94564
4276
交通状况的确比以前糟糕得多。
01:50
And all of this has a very human cost.
33
98840
2409
所有的这一切也都伴随着人力成本。
01:53
So if you take the average commute time
in America, which is about 50 minutes,
in America, which is about 50 minutes,
34
101529
3948
如果你把在美的平均通勤时间，
约50分钟，
约50分钟，
01:57
you multiply that by the 120 million
workers we have,
workers we have,
35
105477
3649
乘以我们的1亿2000万工作者,
02:01
that turns out to be
about six billion minutes
about six billion minutes
36
109126
2225
结果就是60亿分钟，
02:03
wasted in commuting every day.
37
111351
2026
每天会被浪费在路上。
02:05
Now, that's a big number,
so let's put it in perspective.
so let's put it in perspective.
38
113377
2827
这是个很大的数字，
那么我们换个方式,
那么我们换个方式,
02:08
You take that six billion minutes
39
116204
1774
你把这60亿分钟，
02:09
and you divide it by the average
life expectancy of a person,
life expectancy of a person,
40
117978
3784
除以人均寿命，
02:13
that turns out to be 162 lifetimes
41
121762
3135
得出数字是162个生命周期。
02:16
spent every day, wasted,
42
124897
2925
仅仅从A地到B地，每天就有这么多
02:19
just getting from A to B.
43
127822
2044
生命白白浪费掉了。
02:21
It's unbelievable.
44
129866
1730
太难以置信了。
02:23
And then, there are those of us
who don't have the privilege
who don't have the privilege
45
131596
2844
当然，我们当中还有一些人无法享有
02:26
of sitting in traffic.
46
134440
1672
坐到车流中来的权利。
02:28
So this is Steve.
47
136112
1578
这个人叫史蒂夫。
他是个很有才能的人。
02:29
He's an incredibly capable guy,
48
137690
1765
02:31
but he just happens to be blind,
49
139455
2516
但很不幸，他是盲人，
02:33
and that means instead of a 30-minute
drive to work in the morning,
drive to work in the morning,
50
141971
3217
这意味着本来早上
上班路上的30分钟，
上班路上的30分钟，
02:37
it's a two-hour ordeal
of piecing together bits of public transit
of piecing together bits of public transit
51
145188
3979
变成了两个小时的
各种转乘公共交通的折磨，
各种转乘公共交通的折磨，
02:41
or asking friends and family for a ride.
52
149167
2385
或者是请求朋友或家人载他一程。
02:43
He doesn't have that same freedom
that you and I have to get around.
that you and I have to get around.
53
151552
3669
他并没有像你我一样
想去哪儿就去哪儿的自由。
想去哪儿就去哪儿的自由。
02:47
We should do something about that.
54
155221
2460
我们应该改变这种现状。
02:49
Now, conventional wisdom would say
55
157891
1757
现在，传统观点认为,
02:51
that we'll just take
these driver assistance systems
these driver assistance systems
56
159648
2492
我们应该使用驾驶员辅助系统，
02:54
and we'll kind of push them
and incrementally improve them,
and incrementally improve them,
57
162140
3750
然后不停改进它们,
总有一天，它们能够实现自动驾驶。
02:57
and over time, they'll turn
into self-driving cars.
into self-driving cars.
58
165890
2542
03:00
Well, I'm here to tell you
that's like me saying
that's like me saying
59
168432
2409
事实上我今天来就是想告诉你们，
这就跟
这就跟
03:02
that if I work really hard at jumping,
one day I'll be able to fly.
one day I'll be able to fly.
60
170841
4057
如果我努力练习弹跳，
有一天我就能飞翔一样不现实。
有一天我就能飞翔一样不现实。
03:06
We actually need to do
something a little different.
something a little different.
61
174898
2728
我们需要做点不同的东西。
03:09
And so I'm going to talk to you
about three different ways
about three different ways
62
177626
2711
我会跟你们介绍
自动驾驶系统和驾驶员辅助系统的
三个不同方面。
三个不同方面。
03:12
that self-driving systems are different
than driver assistance systems.
than driver assistance systems.
63
180337
3346
03:15
And I'm going to start
with some of our own experience.
with some of our own experience.
64
183683
2651
我先从自己的经历说起。
03:18
So back in 2013,
65
186334
2253
在2013年，
03:20
we had the first test
of a self-driving car
of a self-driving car
66
188587
2663
我们进行了第一次自动驾驶测试，
03:23
where we let regular people use it.
67
191250
2027
让普通人来操作。
03:25
Well, almost regular --
they were 100 Googlers,
they were 100 Googlers,
68
193277
2202
算是普通人吧——
他们是100名谷歌员工，
他们是100名谷歌员工，
03:27
but they weren't working on the project.
69
195479
2003
但他们没有参与开发这个项目。
03:29
And we gave them the car and we allowed
them to use it in their daily lives.
them to use it in their daily lives.
70
197482
3621
我们把车给他们，
让他们在日常生活中使用。
让他们在日常生活中使用。
03:33
But unlike a real self-driving car,
this one had a big asterisk with it:
this one had a big asterisk with it:
71
201103
3719
但与真的自动驾驶汽车不同，
这一辆得加个星号上去：
这一辆得加个星号上去：
03:36
They had to pay attention,
72
204822
1504
他们得留多个心眼儿，
03:38
because this was an experimental vehicle.
73
206326
2633
因为这只是一辆试验车。
03:40
We tested it a lot,
but it could still fail.
but it could still fail.
74
208959
3525
我们虽然进行了很多测试，
但还是有风险。
但还是有风险。
03:44
And so we gave them two hours of training,
75
212484
2059
我们对他们进行了两个小时的训练，
03:46
we put them in the car,
we let them use it,
we let them use it,
76
214543
2092
然后让他们进行实际操作，
03:48
and what we heard back
was something awesome,
was something awesome,
77
216635
2127
然后我们得到了一些很好的反馈，
03:50
as someone trying
to bring a product into the world.
to bring a product into the world.
78
218762
2524
因为有人把产品带到现实中来了。
每个人都对它赞不绝口。
03:53
Every one of them told us they loved it.
79
221286
1925
03:55
In fact, we had a Porsche driver
who came in and told us on the first day,
who came in and told us on the first day,
80
223211
3566
事实上，在第一天
有一个保时捷驾驶员进来跟我们说，
有一个保时捷驾驶员进来跟我们说，
03:58
"This is completely stupid.
What are we thinking?"
What are we thinking?"
81
226777
2663
“这实在是太无厘头了。
你们到底怎么想的？”
你们到底怎么想的？”
04:01
But at the end of it, he said,
"Not only should I have it,
"Not only should I have it,
82
229850
2840
但最后，他说，“不单是我需要它，
每个人都需要有一辆，
大家的车技都太烂了。”
大家的车技都太烂了。”
04:04
everyone else should have it,
because people are terrible drivers."
because people are terrible drivers."
83
232690
3175
04:09
So this was music to our ears,
84
237135
1735
这番话就是我们的福音，
04:10
but then we started to look at what
the people inside the car were doing,
the people inside the car were doing,
85
238870
3803
然后我们开始观察
人们在车里都在做什么，
人们在车里都在做什么，
04:14
and this was eye-opening.
86
242673
1579
真让人大开眼界。
04:16
Now, my favorite story is this gentleman
87
244252
2438
我最喜欢的故事，是一位先生
04:18
who looks down at his phone
and realizes the battery is low,
and realizes the battery is low,
88
246690
3829
低头看手机，发现手机快没电了，
04:22
so he turns around like this in the car
and digs around in his backpack,
and digs around in his backpack,
89
250519
4548
然后他在车里这样转过身，
在背包里四处摸索着，
在背包里四处摸索着，
04:27
pulls out his laptop,
90
255067
2153
拿出他的笔记本电脑，
放到副驾驶座位上，
04:29
puts it on the seat,
91
257220
1567
04:30
goes in the back again,
92
258787
1764
再转过身，
04:32
digs around, pulls out
the charging cable for his phone,
the charging cable for his phone,
93
260551
3367
又摸了一通，拿出手机充电线，
04:35
futzes around, puts it into the laptop,
puts it on the phone.
puts it on the phone.
94
263918
3367
理一下线，插进电脑里，连上手机。
04:39
Sure enough, the phone is charging.
95
267285
2043
棒极了，手机有电了。
04:41
All the time he's been doing
65 miles per hour down the freeway.
65 miles per hour down the freeway.
96
269328
3994
而他那时正行驶在时速65英里的高速上
（约104公里每小时）。
（约104公里每小时）。
04:45
Right? Unbelievable.
97
273322
2484
能想象到吗？太难以置信了。
04:47
So we thought about this and we said,
it's kind of obvious, right?
it's kind of obvious, right?
98
275806
3121
所以我们想了想，说，
这挺明显的对吧？
这挺明显的对吧？
04:50
The better the technology gets,
99
278927
2263
科技越来越发达，
04:53
the less reliable
the driver is going to get.
the driver is going to get.
100
281190
2121
驾驶员就不需要太负责任。
04:55
So by just making the cars
incrementally smarter,
incrementally smarter,
101
283311
2396
所以只是把车变得更加智能，
04:57
we're probably not going to see
the wins we really need.
the wins we really need.
102
285707
2902
并没法让我们看到真正需要的成功。
05:00
Let me talk about something
a little technical for a moment here.
a little technical for a moment here.
103
288609
3901
我在这里要暂时说一点技术上的东西。
05:04
So we're looking at this graph,
and along the bottom
and along the bottom
104
292510
2438
在这张图上，底部的线
05:06
is how often does the car
apply the brakes when it shouldn't.
apply the brakes when it shouldn't.
105
294948
3051
代表着在不必要的时候
制动刹车发生的频率。
制动刹车发生的频率。
05:09
You can ignore most of that axis,
106
297999
1621
你可以忽略这条轴的大部分.
05:11
because if you're driving around town,
and the car starts stopping randomly,
and the car starts stopping randomly,
107
299620
3719
因为如果你在城里开车，
然后车时不时自己停下来。
然后车时不时自己停下来。
05:15
you're never going to buy that car.
108
303339
1701
你永远都不会买这辆车。
05:17
And the vertical axis is how often
the car is going to apply the brakes
the car is going to apply the brakes
109
305040
3375
竖直的轴线表示车会
在你需要避免事故时
在你需要避免事故时
05:20
when it's supposed to
to help you avoid an accident.
to help you avoid an accident.
110
308415
3049
采取制动刹车的频率。
05:23
Now, if we look at
the bottom left corner here,
the bottom left corner here,
111
311464
2221
如果我们看左下角,
05:25
this is your classic car.
112
313685
1845
这是你们正在开的普通汽车。
05:27
It doesn't apply the brakes for you,
it doesn't do anything goofy,
it doesn't do anything goofy,
113
315530
3133
它不会自动为你刹车，
也不至于刹车失灵，
也不至于刹车失灵，
05:30
but it also doesn't get you
out of an accident.
out of an accident.
114
318663
2779
但它无法为你避免事故。
05:33
Now, if we want to bring
a driver assistance system into a car,
a driver assistance system into a car,
115
321442
3018
如果我们把驾驶员辅助系统
装进车里，
装进车里，
比如说撞击缓冲刹车系统，
05:36
say with collision mitigation braking,
116
324460
1828
我们会导入一系列的科技，
05:38
we're going to put some package
of technology on there,
of technology on there,
117
326288
2612
也就是这条曲线，
它有了一些操作属性，
它有了一些操作属性，
05:40
and that's this curve, and it's going
to have some operating properties,
to have some operating properties,
118
328900
3418
但也不会完全规避事故，
05:44
but it's never going to avoid
all of the accidents,
all of the accidents,
119
332318
2490
因为它没有这个能力。
05:46
because it doesn't have that capability.
120
334808
2059
但我们会在曲线上取某个点，
05:48
But we'll pick some place
along the curve here,
along the curve here,
121
336867
2249
也许它可以避免一半
因驾驶员失误引起的事故。
因驾驶员失误引起的事故。
05:51
and maybe it avoids half of accidents
that the human driver misses,
that the human driver misses,
122
339116
3254
挺赞的，对吧？
05:54
and that's amazing, right?
123
342370
1297
我们可以减少一半的交通事故。
05:55
We just reduced accidents on our roads
by a factor of two.
by a factor of two.
124
343667
2727
05:58
There are now 17,000 less people
dying every year in America.
dying every year in America.
125
346394
3987
这样每年在美国就有1万7千人
幸免于难。
幸免于难。
06:02
But if we want a self-driving car,
126
350381
2020
但如果我们想要一辆自动驾驶汽车，
06:04
we need a technology curve
that looks like this.
that looks like this.
127
352401
2307
我们需要一条这样的曲线。
06:06
We're going to have to put
more sensors in the vehicle,
more sensors in the vehicle,
128
354708
2599
我们得在车里加装更多的传感器，
06:09
and we'll pick some
operating point up here
operating point up here
129
357307
2021
然后在这里挑某个操作点，
在这个点上基本
永远不会有事故发生。
永远不会有事故发生。
06:11
where it basically never
gets into a crash.
gets into a crash.
130
359328
2019
06:13
They'll happen, but very low frequency.
131
361347
2443
多少还是会发生，但概率极低。
06:15
Now you and I could look at this
and we could argue
and we could argue
132
363790
2461
现在我们可以看看这里，
探讨一下是否有所增加，
我会提到比方说“80-20规则”，
我会提到比方说“80-20规则”，
06:18
about whether it's incremental, and
I could say something like "80-20 rule,"
I could say something like "80-20 rule,"
133
366251
3605
而且很难再上升了。
06:21
and it's really hard to move up
to that new curve.
to that new curve.
134
369856
2568
但让我们暂时从另一个角度看一看。
06:24
But let's look at it
from a different direction for a moment.
from a different direction for a moment.
135
372424
2934
我们看看这种科技应用的频率多高。
06:27
So let's look at how often
the technology has to do the right thing.
the technology has to do the right thing.
136
375358
3512
06:30
And so this green dot up here
is a driver assistance system.
is a driver assistance system.
137
378870
3506
这个绿点表示驾驶员辅助系统。
06:34
It turns out that human drivers
138
382376
2485
调查发现人类驾驶员
06:36
make mistakes that lead
to traffic accidents
to traffic accidents
139
384861
2647
因为自身错误导致的交通事故，
06:39
about once every 100,000 miles in America.
140
387508
3172
在美国是每10万英里（约16万公里）
发生一次。
发生一次。
06:42
In contrast, a self-driving system
is probably making decisions
is probably making decisions
141
390680
3167
对比之下，自动驾驶系统约在每秒
06:45
about 10 times per second,
142
393847
3663
会自行做10次决定。
所以就数量级而言，
06:49
so order of magnitude,
143
397510
1422
06:50
that's about 1,000 times per mile.
144
398932
2832
约是每英里（1.6公里）1000次。
06:53
So if you compare the distance
between these two,
between these two,
145
401764
2485
如果你对比一下两者的差距，
06:56
it's about 10 to the eighth, right?
146
404249
2600
就是10的八次方，对吧？
06:58
Eight orders of magnitude.
147
406849
1765
8个数量级，
07:00
That's like comparing how fast I run
148
408614
2809
这就像拿我跑步的速度
07:03
to the speed of light.
149
411423
2206
和光速比较，
07:05
It doesn't matter how hard I train,
I'm never actually going to get there.
I'm never actually going to get there.
150
413629
3785
即便我再刻苦训练，
也永远达不到光速。
也永远达不到光速。
07:09
So there's a pretty big gap there.
151
417414
2438
所以这个跨度非常大。
07:11
And then finally, there's how
the system can handle uncertainty.
the system can handle uncertainty.
152
419852
3729
最后，就是这个系统如何
处理不确定性。
处理不确定性。
07:15
So this pedestrian here might be
stepping into the road, might not be.
stepping into the road, might not be.
153
423581
3323
那么这个行人可能会走到路上，
也可能不会。
也可能不会。
07:18
I can't tell,
nor can any of our algorithms,
nor can any of our algorithms,
154
426904
3395
我不确定，也没有任何算法能确定，
但对于驾驶员辅助系统来说，
07:22
but in the case of
a driver assistance system,
a driver assistance system,
155
430310
2284
07:24
that means it can't take action,
because again,
because again,
156
432594
2806
这意味着它无法采取措施，
因为如果它在预期之外采取制动，
是完全不合适的。
是完全不合适的。
07:27
if it presses the brakes unexpectedly,
that's completely unacceptable.
that's completely unacceptable.
157
435400
3339
07:30
Whereas a self-driving system
can look at that pedestrian and say,
can look at that pedestrian and say,
158
438739
3133
但自动驾驶系统则会观察行人，
07:33
I don't know what they're about to do,
159
441872
1890
然后说我不知道他们要做什么，
07:35
slow down, take a better look,
and then react appropriately after that.
and then react appropriately after that.
160
443762
3762
于是减速，再仔细观察，
然后采取适当措施。
然后采取适当措施。
07:39
So it can be much safer than
a driver assistance system can ever be.
a driver assistance system can ever be.
161
447524
3702
所以这就比驾驶员辅助系统
要安全得多，
要安全得多，
07:43
So that's enough about
the differences between the two.
the differences between the two.
162
451226
2730
那么以上的例子就
足以体现这两者的区别了。
足以体现这两者的区别了。
07:45
Let's spend some time talking about
how the car sees the world.
how the car sees the world.
163
453956
3484
现在我们再花点时间聊聊
车是如何观察环境的。
车是如何观察环境的。
这是我们的车。
07:49
So this is our vehicle.
164
457440
1252
07:50
It starts by understanding
where it is in the world,
where it is in the world,
165
458692
2438
它从识别自己的位置开始，
通过将它的地图和
传感器数据进行叠加，
传感器数据进行叠加，
07:53
by taking a map and its sensor data
and aligning the two,
and aligning the two,
166
461130
2787
07:55
and then we layer on top of that
what it sees in the moment.
what it sees in the moment.
167
463917
2948
然后我们再加上它当时看到的东西，
07:58
So here, all the purple boxes you can see
are other vehicles on the road,
are other vehicles on the road,
168
466865
3655
所以在这里所有你能看到的紫色盒子
是路上的其他车辆，
是路上的其他车辆，
08:02
and the red thing on the side
over there is a cyclist,
over there is a cyclist,
169
470520
2528
旁边红色的物体则是一位骑单车的人，
08:05
and up in the distance,
if you look really closely,
if you look really closely,
170
473048
2402
如果你再仔细点看，在远处，
08:07
you can see some cones.
171
475450
1794
你能看到一些锥形路障。
08:09
Then we know where the car
is in the moment,
is in the moment,
172
477244
2773
这样我们就能知道汽车现在的位置了，
08:12
but we have to do better than that:
we have to predict what's going to happen.
we have to predict what's going to happen.
173
480017
3833
但我们还得再改善：
我们得预测将发生的事情。
我们得预测将发生的事情。
这里右上角的小卡车正准备换到左道,
08:15
So here the pickup truck in top right
is about to make a left lane change
is about to make a left lane change
174
483850
3488
因为前面的路被封了，
08:19
because the road in front of it is closed,
175
487338
2223
所以它得驶离原车道。
08:21
so it needs to get out of the way.
176
489561
1731
08:23
Knowing that one pickup truck is great,
177
491292
1863
能知道一辆小卡车的轨迹确实不错,
08:25
but we really need to know
what everybody's thinking,
what everybody's thinking,
178
493155
2479
但我们真正需要的是
了解每个人的想法,
了解每个人的想法,
08:27
so it becomes quite a complicated problem.
179
495634
2507
所以问题就变得十分复杂了。
08:30
And then given that, we can figure out
how the car should respond in the moment,
how the car should respond in the moment,
180
498141
4749
了解这些之后，
我们就可以算出汽车该如何应对，
我们就可以算出汽车该如何应对，
该跟随哪条线路，
要多快实现减速或者加速。
要多快实现减速或者加速。
08:34
so what trajectory it should follow, how
quickly it should slow down or speed up.
quickly it should slow down or speed up.
181
502890
3866
08:38
And then that all turns into
just following a path:
just following a path:
182
506756
3065
所有的这一切最终都会
变成跟随一条路径：
变成跟随一条路径：
08:41
turning the steering wheel left or right,
pressing the brake or gas.
pressing the brake or gas.
183
509821
3197
向左或还是右打方向盘，
踩刹车还是油门。
踩刹车还是油门。
08:45
It's really just two numbers
at the end of the day.
at the end of the day.
184
513018
2464
所有的一切最终都
简化成了两个数值，
简化成了两个数值，
08:47
So how hard can it really be?
185
515482
2241
那这能有多难呢？
08:50
Back when we started in 2009,
186
518433
1952
在2009年我们刚开始测试的时候，
08:52
this is what our system looked like.
187
520385
1798
我们的系统看起来是这样的。
你能看到我们的车在中间，
路上还有其他盒子，
路上还有其他盒子，
08:54
So you can see our car in the middle
and the other boxes on the road,
and the other boxes on the road,
188
522183
3391
同时在高速上行驶着。
08:57
driving down the highway.
189
525574
1271
这辆车需要知道自己的位置
以及其他车辆的大概方位。
以及其他车辆的大概方位。
08:58
The car needs to understand where it is
and roughly where the other vehicles are.
and roughly where the other vehicles are.
190
526845
3818
基本上就是一种几何的分析方式。
09:02
It's really a geometric
understanding of the world.
understanding of the world.
191
530663
2429
09:05
Once we started driving
on neighborhood and city streets,
on neighborhood and city streets,
192
533092
2948
当我们开始在社区和
城市道路上行驶时，
城市道路上行驶时，
09:08
the problem becomes a whole
new level of difficulty.
new level of difficulty.
193
536040
2445
问题又上升到了一个新的难度。
09:10
You see pedestrians crossing in front
of us, cars crossing in front of us,
of us, cars crossing in front of us,
194
538485
3494
你能看到行人在我们面前穿梭，
还有汽车,
还有汽车,
09:13
going every which way,
195
541979
1811
横纵交错，
还有红绿灯，人行横道。
09:15
the traffic lights, crosswalks.
196
543790
1527
09:17
It's an incredibly complicated
problem by comparison.
problem by comparison.
197
545317
2797
相比之下问题变得极度复杂。
09:20
And then once you have
that problem solved,
that problem solved,
198
548114
2103
当你把这个问题解决后，
09:22
the vehicle has to be able
to deal with construction.
to deal with construction.
199
550217
2512
接下来汽车还得能应付建筑施工。
09:24
So here are the cones on the left
forcing it to drive to the right,
forcing it to drive to the right,
200
552729
3151
所以左边的锥形路障会
迫使汽车开到右边，
迫使汽车开到右边，
09:27
but not just construction
in isolation, of course.
in isolation, of course.
201
555880
2402
当然不仅需要避开施工区域，
09:30
It has to deal with other people moving
through that construction zone as well.
through that construction zone as well.
202
558282
3723
它还得应付在其间走动的其他人。
09:34
And of course, if anyone's
breaking the rules, the police are there
breaking the rules, the police are there
203
562005
3263
当然，如果有人违规了，
有警察在那里，
有警察在那里，
汽车就得明白车上闪着灯
09:37
and the car has to understand that
that flashing light on the top of the car
that flashing light on the top of the car
204
565268
3622
意味着这不仅仅是一辆车，
还有一位警官。
还有一位警官。
09:40
means that it's not just a car,
it's actually a police officer.
it's actually a police officer.
205
568890
3105
类似的，这里路边橘黄色的盒子，
09:43
Similarly, the orange box
on the side here,
on the side here,
206
571995
2032
是校车，
09:46
it's a school bus,
207
574027
1109
09:47
and we have to treat that
differently as well.
differently as well.
208
575136
2520
我们也得分开来处理。
09:50
When we're out on the road,
other people have expectations:
other people have expectations:
209
578576
2793
当我们在路上时，
其他人会表达各种意图：
其他人会表达各种意图：
09:53
So, when a cyclist puts up their arm,
210
581369
1780
当骑单车的人举起他们的手臂，
09:55
it means they're expecting the car
to yield to them and make room for them
to yield to them and make room for them
211
583149
3518
这就意味着他们希望有车能
让给他们点空间
让给他们点空间
09:58
to make a lane change.
212
586667
2053
以便进行换道。
10:01
And when a police officer
stood in the road,
stood in the road,
213
589030
2173
当路中间站着一位警官，
10:03
our vehicle should understand
that this means stop,
that this means stop,
214
591203
2740
我们的车得明白这手势是要你停下来，
10:05
and when they signal to go,
we should continue.
we should continue.
215
593943
3506
当他们示意我们可以走了，
我们才能继续。
我们才能继续。
10:09
Now, the way we accomplish this
is by sharing data between the vehicles.
is by sharing data between the vehicles.
216
597449
3761
我们达成这些目标，
是通过和其他车辆分享数据。
是通过和其他车辆分享数据。
10:13
The first, most crude model of this
217
601210
1696
首先，最简单粗制的模型,
10:14
is when one vehicle
sees a construction zone,
sees a construction zone,
218
602906
2113
就是当一辆车看到建筑施工地带时,
10:17
having another know about it
so it can be in the correct lane
so it can be in the correct lane
219
605019
3062
告知另一辆车，让它驶上正确的车道
10:20
to avoid some of the difficulty.
220
608081
1570
以避免麻烦。
10:21
But we actually have a much
deeper understanding of this.
deeper understanding of this.
221
609651
2664
但我们对此有更深的认识。
10:24
We could take all of the data
that the cars have seen over time,
that the cars have seen over time,
222
612315
3009
我们可以搜集车辆在一段时间内
看到的数据，
看到的数据，
10:27
the hundreds of thousands
of pedestrians, cyclists,
of pedestrians, cyclists,
223
615324
2376
数以千计在路上的行人，骑单车的人，
10:29
and vehicles that have been out there
224
617700
1787
以及其他车辆，
10:31
and understand what they look like
225
619487
1695
了解他们的外形，
10:33
and use that to infer
what other vehicles should look like
what other vehicles should look like
226
621182
2831
再用之去推理其他车辆
以及其他行人的外形。
10:36
and other pedestrians should look like.
227
624013
1926
10:37
And then, even more importantly,
we could take from that a model
we could take from that a model
228
625939
3021
然后，更重要的是，
我们可以从中得出一个模型,
我们可以从中得出一个模型,
10:40
of how we expect them
to move through the world.
to move through the world.
229
628960
2330
预测所有交通参与者的去向。
10:43
So here the yellow box is a pedestrian
crossing in front of us.
crossing in front of us.
230
631290
2963
这里的黄色盒子是
我们面前过马路的行人。
我们面前过马路的行人。
10:46
Here the blue box is a cyclist
and we anticipate
and we anticipate
231
634253
2250
这个蓝色盒子是个骑单车的人，
而且我们预测
而且我们预测
10:48
that they're going to nudge out
and around the car to the right.
and around the car to the right.
232
636503
3312
他会一直保持在车辆右侧。
10:52
Here there's a cyclist
coming down the road
coming down the road
233
640115
2092
这里有个在路上骑单车的人，
10:54
and we know they're going to continue
to drive down the shape of the road.
to drive down the shape of the road.
234
642207
3486
我们知道他会沿着路一直骑下去。
这里有人右转了,
10:57
Here somebody makes a right turn,
235
645693
1867
而在这里，有人会在我们面前调头，
10:59
and in a moment here, somebody's
going to make a U-turn in front of us,
going to make a U-turn in front of us,
236
647560
3360
我们可以预测这个行为并安全应对。
11:02
and we can anticipate that behavior
and respond safely.
and respond safely.
237
650920
2614
目前为止，对于我们见过的场景
都没什么问题，
都没什么问题，
11:05
Now, that's all well and good
for things that we've seen,
for things that we've seen,
238
653534
2728
但当然，你还会遇见很多
11:08
but of course, you encounter
lots of things that you haven't
lots of things that you haven't
239
656262
2865
之前没见过的东西。
11:11
seen in the world before.
240
659127
1231
几个月前，
11:12
And so just a couple of months ago,
241
660358
1741
我们的车辆在通过Mountain View
（硅谷地名）的时候，
（硅谷地名）的时候，
11:14
our vehicles were driving
through Mountain View,
through Mountain View,
242
662099
2235
就遇到了这样的情景。
11:16
and this is what we encountered.
243
664334
1644
这是个坐着电动轮椅的女士,
11:17
This is a woman in an electric wheelchair
244
665978
2082
在路中央绕着圈追赶一只鸭子。
（笑声）
（笑声）
11:20
chasing a duck in circles on the road.
(Laughter)
(Laughter)
245
668060
2617
11:22
Now it turns out, there is nowhere
in the DMV handbook
in the DMV handbook
246
670677
3111
结果呢，在机动车驾驶管理处的手册里
11:25
that tells you how to deal with that,
247
673788
2245
没有一条告诉你该怎么做，
11:28
but our vehicles were able
to encounter that,
to encounter that,
248
676033
2143
但我们的车辆却能灵活应对，
减速并安全通过。
11:30
slow down, and drive safely.
249
678176
2255
11:32
Now, we don't have to deal
with just ducks.
with just ducks.
250
680431
2041
我们应付的不只是鸭子,
看看这只飞过我们面前的鸟，
汽车也会对之做出处理。
汽车也会对之做出处理。
11:34
Watch this bird fly across in front of us.
The car reacts to that.
The car reacts to that.
251
682472
3708
11:38
Here we're dealing with a cyclist
252
686180
1615
这里还有一个骑车的,
11:39
that you would never expect to see
anywhere other than Mountain View.
anywhere other than Mountain View.
253
687795
3290
估计除了在Mountain View，
其他地方很难见到。
其他地方很难见到。
11:43
And of course, we have
to deal with drivers,
to deal with drivers,
254
691085
2068
当然，我们还得应付其他驾驶员，
11:45
even the very small ones.
255
693153
3715
甚至那些幼龄的。
11:48
Watch to the right as someone
jumps out of this truck at us.
jumps out of this truck at us.
256
696868
4131
注意右边，那个从货车上跳下来的人。
现在，注意左边绿盒子那辆车，
11:54
And now, watch the left as the car
with the green box decides
with the green box decides
257
702460
2929
11:57
he needs to make a right turn
at the last possible moment.
at the last possible moment.
258
705389
3325
它决定在最后的时刻右转。
12:00
Here, as we make a lane change,
the car to our left decides
the car to our left decides
259
708714
2851
这里，当我们变道时，我们左边的车
也同样想变道。
12:03
it wants to as well.
260
711565
3553
12:07
And here, we watch a car
blow through a red light
blow through a red light
261
715118
2693
还有这里，我们看到一辆车闯了红灯，
12:09
and yield to it.
262
717811
2090
我们就先让它通过。
12:11
And similarly, here, a cyclist
blowing through that light as well.
blowing through that light as well.
263
719901
3854
同样这里，骑单车的人也闯红灯了，
12:15
And of course,
the vehicle responds safely.
the vehicle responds safely.
264
723755
2746
不出所料，
我们的车也能安全应对。
我们的车也能安全应对。
12:18
And of course, we have people
who do I don't know what
who do I don't know what
265
726501
2601
当然还有一些莫名其妙的人，
12:21
sometimes on the road, like this guy
pulling out between two self-driving cars.
pulling out between two self-driving cars.
266
729102
3823
就像这家伙一样，
直接就从两辆自动驾驶汽车中窜出来。
直接就从两辆自动驾驶汽车中窜出来。
你会想问，“你脑子是怎么想的？”
12:24
You have to ask, "What are you thinking?"
267
732925
2045
12:26
(Laughter)
268
734970
1212
（笑声）
12:28
Now, I just fire-hosed you
with a lot of stuff there,
with a lot of stuff there,
269
736182
2521
我已经给你们看了大量的例子，
我快速地讲一下其中的一个案例。
12:30
so I'm going to break one of these
down pretty quickly.
down pretty quickly.
270
738703
2650
我们现在看到的还是骑单车的人，
12:33
So what we're looking at is the scene
with the cyclist again,
with the cyclist again,
271
741353
2940
你们可能会发现在下面的视角，
我们还看不到那个人，
我们还看不到那个人，
12:36
and you might notice in the bottom,
we can't actually see the cyclist yet,
we can't actually see the cyclist yet,
272
744293
3491
但车能看到：就是那里的小蓝盒子,
12:39
but the car can: it's that little
blue box up there,
blue box up there,
273
747784
2504
这来自于激光数据。
12:42
and that comes from the laser data.
274
750288
2081
这并不是很容易理解，
12:44
And that's not actually
really easy to understand,
really easy to understand,
275
752369
2418
那么我接下来要做的，
就是调出激光数据看一下,
就是调出激光数据看一下,
12:46
so what I'm going to do is I'm going
to turn that laser data and look at it,
to turn that laser data and look at it,
276
754787
3584
如果你擅长分析激光数据，你能发现
12:50
and if you're really good at looking
at laser data, you can see
at laser data, you can see
277
758371
3029
曲线上的一些点，
就在那儿，
那个蓝色小盒子就是骑单车的人。
那个蓝色小盒子就是骑单车的人。
12:53
a few dots on the curve there,
278
761400
1487
12:54
right there, and that blue box
is that cyclist.
is that cyclist.
279
762887
2372
这会儿面对我们的还是红灯，
12:57
Now as our light is red,
280
765259
1149
自行车道的灯已经变黄了，
12:58
the cyclist's light
has turned yellow already,
has turned yellow already,
281
766408
2192
如果你瞥一眼的话，就能看到了。
13:00
and if you squint, you can see that
in the imagery.
in the imagery.
282
768600
2438
但是骑单车的人，
我们看到他准备穿过这个十字路口，
我们看到他准备穿过这个十字路口，
13:03
But the cyclist, we see, is going
to proceed through the intersection.
to proceed through the intersection.
283
771038
3286
我们的灯已经变绿，
他的方向也变红了，
他的方向也变红了，
13:06
Our light has now turned green,
his is solidly red,
his is solidly red,
284
774324
2394
13:08
and we now anticipate that this bike
is going to come all the way across.
is going to come all the way across.
285
776718
4292
我们预期到这辆单车将会横穿马路。
13:13
Unfortunately the other drivers next to us
were not paying as much attention.
were not paying as much attention.
286
781010
3742
但不幸的是，
我们旁边的其他司机并没有注意到。
我们旁边的其他司机并没有注意到。
13:16
They started to pull forward,
and fortunately for everyone,
and fortunately for everyone,
287
784752
3157
他们开始踩油门,
不过幸运的是，
不过幸运的是，
13:19
this cyclists reacts, avoids,
288
787909
3011
骑单车的人及时避开了，
然后平安地穿过了十字路口。
13:22
and makes it through the intersection.
289
790920
2191
13:25
And off we go.
290
793111
1568
之后我们才又继续前进。
13:26
Now, as you can see, we've made
some pretty exciting progress,
some pretty exciting progress,
291
794679
2948
正如你们所见，
我们已取得了一些激动人心的成就，
我们已取得了一些激动人心的成就，
13:29
and at this point we're pretty convinced
292
797627
1902
此时我们深信
13:31
this technology is going
to come to market.
to come to market.
293
799529
2010
这项技术将会进入市场。
13:33
We do three million miles of testing
in our simulators every single day,
in our simulators every single day,
294
801539
4783
我们每天用虚拟器做
3百万英里的测试，
3百万英里的测试，
13:38
so you can imagine the experience
that our vehicles have.
that our vehicles have.
295
806322
2689
所以你就能够想象到
我们的车辆获得了多少经验。
我们的车辆获得了多少经验。
13:41
We are looking forward to having
this technology on the road,
this technology on the road,
296
809011
2864
我们期待这项技术能在道路上使用，
13:43
and we think the right path
is to go through the self-driving
is to go through the self-driving
297
811875
2890
并且认为正确的选择是自动驾驶，
13:46
rather than driver assistance approach
298
814765
1844
而非驾驶员辅助系统，
13:48
because the urgency is so large.
299
816609
2621
因为情况已经刻不容缓了。
13:51
In the time I have given this talk today,
300
819230
2393
就在我演讲的时间段内,
13:53
34 people have died on America's roads.
301
821623
3135
在美国的公路上已经有34人丧生。
13:56
How soon can we bring it out?
302
824758
2368
我们多久能实现这个目标呢？
13:59
Well, it's hard to say because
it's a really complicated problem,
it's a really complicated problem,
303
827126
3832
嗯，很难说，
因为这是个很复杂的问题，
因为这是个很复杂的问题，
14:02
but these are my two boys.
304
830958
2214
这两个是我的儿子。
14:05
My oldest son is 11, and that means
in four and a half years,
in four and a half years,
305
833172
3623
大的11岁，也就是说在四年半后，
14:08
he's going to be able
to get his driver's license.
to get his driver's license.
306
836795
2577
他就能去考驾照了。
14:11
My team and I are committed
to making sure that doesn't happen.
to making sure that doesn't happen.
307
839372
3204
我和我的团队承诺
尽量不让他去考（已经不需要了）。
尽量不让他去考（已经不需要了）。
14:14
Thank you.
308
842576
1904
谢谢。
14:16
(Laughter) (Applause)
309
844480
3667
（笑声）（掌声）
克利斯·安德森（CA）：
克里斯，我有个问题要问你。
克里斯，我有个问题要问你。
14:21
Chris Anderson: Chris,
I've got a question for you.
I've got a question for you.
310
849110
2568
克里斯·厄姆森（CU）：问吧。
14:23
Chris Urmson: Sure.
311
851678
2809
14:26
CA: So certainly, the mind of your cars
is pretty mind-boggling.
is pretty mind-boggling.
312
854487
3924
CA：显而易见，
你们的车有着让人惊奇的大脑。
你们的车有着让人惊奇的大脑。
14:30
On this debate between
driver-assisted and fully driverless --
driver-assisted and fully driverless --
313
858411
4459
在驾驶辅助和无人驾驶这场辩论上——
14:34
I mean, there's a real debate
going on out there right now.
going on out there right now.
314
862870
3041
我是说，现在就有一场真正的辩论。
14:37
So some of the companies,
for example, Tesla,
for example, Tesla,
315
865911
2833
一些公司，例如，特斯拉，
14:40
are going the driver-assisted route.
316
868744
2159
正在走驾驶辅助的路线。
14:42
What you're saying is that
that's kind of going to be a dead end
that's kind of going to be a dead end
317
870903
5248
而你所说的，
就是这将是个没前途的死胡同,
就是这将是个没前途的死胡同,
14:48
because you can't just keep improving
that route and get to fully driverless
that route and get to fully driverless
318
876151
5456
因为你不能指望在这方面不断提高
就会在某个时候实现无人驾驶，
就会在某个时候实现无人驾驶，
14:53
at some point, and then a driver
is going to say, "This feels safe,"
is going to say, "This feels safe,"
319
881607
3530
然后有驾驶员说
“这已经挺安全的了”，
“这已经挺安全的了”，
14:57
and climb into the back,
and something ugly will happen.
and something ugly will happen.
320
885137
2647
然后转身去后座，
不幸就发生了。
不幸就发生了。
14:59
CU: Right. No, that's exactly right,
and it's not to say
and it's not to say
321
887784
2676
CU：对，你说得对，这并不是说
驾驶员辅助系统作用不大。
15:02
that the driver assistance systems
aren't going to be incredibly valuable.
aren't going to be incredibly valuable.
322
890460
3537
它能在这个过渡阶段拯救很多生命,
15:05
They can save a lot of lives
in the interim,
in the interim,
323
893997
2058
但为了抓住这一变革性的机会，
能帮助像史蒂夫这样的人行动自如，
能帮助像史蒂夫这样的人行动自如，
15:08
but to see the transformative opportunity
to help someone like Steve get around,
to help someone like Steve get around,
324
896055
3833
为了终结安全事故，
15:11
to really get to the end case in safety,
325
899888
1969
15:13
to have the opportunity
to change our cities
to change our cities
326
901857
2479
为了有机会改变我们的城市，
15:16
and move parking out and get rid of
these urban craters we call parking lots,
these urban craters we call parking lots,
327
904336
4204
解决停车问题，
摆脱我们称为停车场的城市大坑,
摆脱我们称为停车场的城市大坑,
这是唯一的办法了。
15:20
it's the only way to go.
328
908540
1240
15:21
CA: We will be tracking your progress
with huge interest.
with huge interest.
329
909780
2718
CA：我们会带着浓厚的兴趣
关注你们的进展的。
关注你们的进展的。
15:24
Thanks so much, Chris.
CU: Thank you. (Applause)
CU: Thank you. (Applause)
330
912498
4232
谢谢你，克里斯。
CU：谢谢。（掌声）
CU：谢谢。（掌声）
ABOUT THE SPEAKER
Chris Urmson - RoboticistChris Umson is the Director of Self-Driving Cars at Google[x].
Why you should listen
Since 2009, Chris Urmson has headed up Google’s self-driving car program. So far, the team’s vehicles have driven over three quarters of a million miles. While early models included a driverless Prius that TEDsters got to test- ... um, -not-drive in 2011, more and more the team is building vehicles from the ground up, custom-made to go driverless.
Prior to joining Google, Umson was on the faculty of the Robotics Institute at Carnegie Mellon University, where his research focused on motion planning and perception for robotic vehicles. During his time at Carnegie Mellon, he served as Director of Technology for the team that won the 2007 DARPA Urban Challenge.
Chris Urmson | Speaker | TED.com