9 - 8 - Autonomous Driving (7 min).srt

1
00:00:00,090 --> 00:00:01,100
In this video, I'd like to
在这段视频中
(字幕整理：中国海洋大学 黄海广,haiguang2000@qq.com )

2
00:00:01,200 --> 00:00:02,840
show you a fun and historically
我想向你介绍一个具有历史意义的

3
00:00:03,390 --> 00:00:05,820
important example of Neural Network Learning.
神经网络学习的重要例子

4
00:00:06,720 --> 00:00:09,300
Of using a Neural Network for autonomous driving
那就是使用神经网络来实现自动驾驶

5
00:00:09,870 --> 00:00:12,430
that is getting a car to learn to drive itself.
也就是说使汽车通过学习来自己驾驶

6
00:00:13,810 --> 00:00:14,980
The video that I
接下来我将演示的

7
00:00:15,130 --> 00:00:16,450
showed a minute, was something
这段视频

8
00:00:16,820 --> 00:00:18,290
that I've gotten from Dean Pomilieu,
是我从 Dean Pomerleau那里拿到的

9
00:00:18,470 --> 00:00:20,000
who Colleague who works
他是我的同事

10
00:00:20,260 --> 00:00:22,000
out in Carnegie Mellon University out
任职于美国东海岸的

11
00:00:22,140 --> 00:00:23,440
on the east coast of the United States,
卡耐基梅隆大学

12
00:00:24,460 --> 00:00:25,310
and in part of the
在这部分视频中

13
00:00:25,350 --> 00:00:27,980
video you see visualizations like
你就会明白可视化技术到底是什么

14
00:00:28,230 --> 00:00:29,930
this, and what I should tell you what the visualization
在看这段视频之前

15
00:00:30,080 --> 00:00:31,170
looks like before starting to
我会告诉你可视化技术是什么

16
00:00:31,260 --> 00:00:32,830
video. Down here
在下面

17
00:00:33,170 --> 00:00:34,860
on the lower left is the
也就是左下方

18
00:00:35,100 --> 00:00:36,150
view seen by the car
就是汽车所看到的

19
00:00:36,750 --> 00:00:37,580
of what's in front of it
前方的路况图像

20
00:00:37,840 --> 00:00:38,980
and so here you know, you will kind
在图中你依稀能看出

21
00:00:39,130 --> 00:00:40,250
of see you know, a road that's
一条道路

22
00:00:40,450 --> 00:00:41,390
maybe going a bit to
朝左延伸了一点

23
00:00:41,470 --> 00:00:42,100
the left and going a little bit to
又向右了一点

24
00:00:42,670 --> 00:00:45,060
the right, and up
然后上面的这幅图

25
00:00:45,250 --> 00:00:47,230
here on top, this
你可以看到一条

26
00:00:47,820 --> 00:00:49,820
first horizontal bar shows the
水平的菜单栏

27
00:00:49,940 --> 00:00:51,510
direction selected by the
显示的是驾驶操作人

28
00:00:51,750 --> 00:00:53,110
human driver and is the
所选择的方向

29
00:00:53,580 --> 00:00:54,630
location of this bright
就是这里的

30
00:00:55,180 --> 00:00:56,830
white band that shows the
这条白亮的区段

31
00:00:57,100 --> 00:00:58,490
steering direction selected by the
显示的就是

32
00:00:58,690 --> 00:01:00,450
human driver, where, you
人类驾驶者选择的方向

33
00:01:00,680 --> 00:01:01,780
know, here, far to the left
比如 最左边的区段

34
00:01:02,150 --> 00:01:03,280
corresponds to steering hard left;
对应的操作就是向左急转

35
00:01:03,910 --> 00:01:05,180
here corresponds to steering hard
而最右端则对应

36
00:01:05,450 --> 00:01:06,830
to the right; and so
向右急转的操作

37
00:01:06,980 --> 00:01:08,280
this location, which is a
因此 稍微靠左的区段

38
00:01:08,500 --> 00:01:09,390
little bit to the left,
也就是这里

39
00:01:09,720 --> 00:01:10,730
a little bit left of
中心稍微向左一点的位置

40
00:01:10,890 --> 00:01:12,120
center, means that the human
则表示在这一点上

41
00:01:12,280 --> 00:01:13,350
driver, at this point, was
人类驾驶者的操作是

42
00:01:13,520 --> 00:01:14,600
steering slightly to the left.  A
慢慢的向左拐

43
00:01:16,020 --> 00:01:17,340
nd this second part here
这幅图的第二部分

44
00:01:17,880 --> 00:01:18,800
corresponds to the steering
对应的就是

45
00:01:19,140 --> 00:01:20,720
direction selected by the
学习算法选出的行驶方向

46
00:01:20,870 --> 00:01:22,020
learning algorithm; and again, the
并且 类似的

47
00:01:22,110 --> 00:01:23,060
location of this sort
这一条白亮的区段

48
00:01:23,310 --> 00:01:24,790
of white band, means the
显示的就是

49
00:01:24,850 --> 00:01:26,560
neural network was here, selecting
神经网络在这里

50
00:01:27,040 --> 00:01:28,300
a steering direction just slightly to
选择的行驶方向

51
00:01:28,380 --> 00:01:29,440
the left and in fact,
是稍微的左转 并且实际上

52
00:01:29,970 --> 00:01:30,980
before the neural network starts
在神经网络开始

53
00:01:31,390 --> 00:01:33,020
learning initially, you
学习之前

54
00:01:33,170 --> 00:01:34,990
see that the network outputs a
你会看到网络的输出是

55
00:01:35,170 --> 00:01:36,410
grey band, like a
一条灰色的区段

56
00:01:36,580 --> 00:01:38,500
grey uniform, grey band throughout
就像这样的一条灰色区段

57
00:01:38,890 --> 00:01:40,260
this region, so the uniform
覆盖着整个区域  这些均称的

58
00:01:40,740 --> 00:01:42,210
grey fuzz corresponds to the
灰色区域显示出

59
00:01:42,330 --> 00:01:43,880
neural network having been randomly
神经网络已经随机初始化了

60
00:01:44,450 --> 00:01:46,180
initialized, and initially having
并且初始化时

61
00:01:46,510 --> 00:01:47,960
no idea how to
我们并不知道 汽车如何行驶

62
00:01:48,020 --> 00:01:49,650
drive the car, or initially having
或者说

63
00:01:50,050 --> 00:01:52,500
no idea what direction to steer in.
我们并不知道所选行驶方向

64
00:01:52,590 --> 00:01:53,640
And it's only after it's learned
只有在学习算法运行了

65
00:01:53,860 --> 00:01:55,290
for a while that it will then start
足够长的时间之后

66
00:01:55,700 --> 00:01:57,520
to output like a solid white
才会有这条白色的区段

67
00:01:57,770 --> 00:01:58,640
band in just a small
出现在整条灰色区域之中

68
00:01:58,800 --> 00:02:00,260
part of the region corresponding
显示出一个具体的

69
00:02:00,700 --> 00:02:01,870
to choosing a particular steering direction.
行驶方向

70
00:02:02,960 --> 00:02:04,710
And that corresponds to when a neural network.
这就表示神经网络算法

71
00:02:05,340 --> 00:02:06,890
Becomes more confident in selecting, you know,
在这时候已经选出了一个

72
00:02:08,080 --> 00:02:09,250
a and in one
明确的行驶方向

73
00:02:10,220 --> 00:02:11,560
location rather than outputting
不像刚开始的时候

74
00:02:11,920 --> 00:02:13,110
a sort of light gray
输出一段模糊的浅灰色区域

75
00:02:13,300 --> 00:02:14,570
fuzz, but instead outputting
而是输出

76
00:02:14,970 --> 00:02:17,010
a white band that's
一条白亮的区段

77
00:02:17,410 --> 00:02:19,220
more constantly selecting one steering direction.
表示已经选出了明确的行驶方向

78
00:02:21,340 --> 00:02:21,880
Alban is a system
ALVINN (Autonomous Land Vehicle In a Neural Network)

79
00:02:22,340 --> 00:02:24,850
of artificial neural networks, that learns to steer
是一个基于神经网络的智能系统

80
00:02:25,280 --> 00:02:26,400
by watching a person drive. Alban
通过观察人类的驾驶来学习驾驶

81
00:02:27,590 --> 00:02:29,550
is designed to control the
ALVINN能够控制NavLab载具——

82
00:02:29,670 --> 00:02:31,310
tube a modified army
一辆改装版军用悍马

83
00:02:31,920 --> 00:02:32,840
Humvee who could put
这辆悍马装载了

84
00:02:33,020 --> 00:02:35,200
sensors, computers and actuators
传感器 计算机和驱动器

85
00:02:36,160 --> 00:02:37,800
for autonomous navigation experiments.
用来进行自动驾驶的导航试验

86
00:02:41,190 --> 00:02:42,480
The initial spec in
实现ALVINN功能的第一步

87
00:02:42,750 --> 00:02:44,730
configuring Alban is training in
是对它进行训练

88
00:02:46,770 --> 00:02:48,160
the training the person drives
也就是训练一个人驾驶汽车

89
00:02:48,580 --> 00:02:50,640
to be a car while Alban watches.
然后让ALVINN观看

90
00:02:55,810 --> 00:02:58,420
Once every two seconds, Alban
ALVINN每两秒

91
00:02:58,690 --> 00:02:59,800
digitizes a video image
将前方的路况图生成一张数字化图片

92
00:03:00,320 --> 00:03:03,260
of the road ahead, and records the person's steering direction.
并且记录驾驶者的驾驶方向

93
00:03:11,790 --> 00:03:13,140
This training image is reduced
得到的训练集图片

94
00:03:13,560 --> 00:03:15,260
in resolution to 30 by
被压缩为30x32像素

95
00:03:15,470 --> 00:03:16,980
32 pixels and provided
并且作为输入

96
00:03:17,590 --> 00:03:19,100
as input to Alban's three-layer
提供给ALVINN的三层

97
00:03:21,920 --> 00:03:21,920
network.
神经网络

98
00:03:22,460 --> 00:03:25,330
Using the back propagation learning algorithm; Alban
通过使用反向传播学习算法

99
00:03:25,370 --> 00:03:26,590
is training to output the same
ALVINN会训练得到一个

100
00:03:26,940 --> 00:03:28,450
steering direction as the
与人类驾驶员操纵方向

101
00:03:28,530 --> 00:03:29,970
human driver for that image
基本相近的结果

102
00:03:33,450 --> 00:03:35,970
Initially, the network's steering response is random.
一开始 我们的网络选择出的方向是随机的

103
00:03:43,930 --> 00:03:45,010
After about two minutes of
大约经过两分钟的训练后

104
00:03:45,100 --> 00:03:46,760
training, the network learns
我们的神经网络

105
00:03:47,080 --> 00:03:48,730
to accurately imitate the steering
便能够准确地模拟

106
00:03:49,110 --> 00:03:56,430
reactions of the
人类驾驶者的

107
00:03:58,670 --> 00:04:03,440
human driver. This same
驾驶方向

108
00:04:03,710 --> 00:04:06,440
training procedure is repeated for other road types.
对其他道路类型 也重复进行这个训练过程

109
00:04:09,940 --> 00:04:11,680
After the networks have been trained the
当网络被训练完成后

110
00:04:11,770 --> 00:04:12,900
operator pushes the run
操作者就可按下运行按钮

111
00:04:13,200 --> 00:04:14,650
switch and often begins
车辆便开始行驶

112
00:04:15,050 --> 00:04:20,380
driving. 12 times
每秒钟ALVINN生成

113
00:04:20,870 --> 00:04:23,010
per second, Alban digitizes an
12次数字化图片

114
00:04:23,090 --> 00:04:25,130
image and feeds it to its neural networks.
并且将图像传送给神经网络进行训练

115
00:04:33,210 --> 00:04:34,980
Each network, running in parallel,
多个神经网络同时工作

116
00:04:35,930 --> 00:04:38,140
produces a steering direction and a measure of it's
每一个网络都生成一个行驶方向

117
00:04:40,050 --> 00:04:40,050
confidence in its response.
以及一个预测自信度的参数

118
00:04:46,610 --> 00:04:49,480
The steering direction
预测自信度最高的

119
00:04:50,140 --> 00:04:51,120
from the most confident network.
那个神经网络得到的行驶方向

120
00:04:52,340 --> 00:04:53,650
In this case, the network trained
比如这里 在这条单行道上训练出的网络

121
00:04:54,000 --> 00:04:56,640
for the one-lane road is used to control the vehicle.
将被最终用于控制车辆方向

122
00:05:04,750 --> 00:05:07,840
Suddenly, an intersection appears ahead
车辆前方突然出现了

123
00:05:08,310 --> 00:05:09,350
of the vehicle.
一个交叉十字路口

124
00:05:23,090 --> 00:05:24,450
As the vehicle approaches the intersection,
当车辆到达这个十字路口时

125
00:05:25,680 --> 00:05:27,740
the confidence of the one-lane network decreases.
我们单行道网络对应的自信度骤减

126
00:05:38,070 --> 00:05:40,030
As it crosses the intersection, and
当它穿过这个十字路口时

127
00:05:40,130 --> 00:05:41,160
the two-lane road ahead comes
前方的双车道将进入其视线

128
00:05:41,440 --> 00:05:44,610
into view, the confidence of the two-lane network rises.
双车道网络的自信度便开始上升

129
00:05:51,260 --> 00:05:53,000
When it's confidence rises, the two-lane
当它的自信度上升时 双车道的网络

130
00:05:53,420 --> 00:05:54,630
network is selected to steer,
将被选择来控制行驶方向

131
00:05:55,050 --> 00:05:56,780
safely guiding the vehicle
车辆将被安全地引导

132
00:05:57,380 --> 00:05:59,030
into it's lane, on the two-lane road.
进入双车道路

133
00:06:05,400 --> 00:06:06,670
So that was autonomous
这就是基于神经网络的

134
00:06:07,010 --> 00:06:09,790
driving using a neural network. Of course, there are more
自动驾驶技术 当然 我们还有很多

135
00:06:10,150 --> 00:06:11,070
recently more modern attempts
更加先进的试验

136
00:06:11,910 --> 00:06:14,000
to do autonomous driving in a few properties, in
来实现自动驾驶技术

137
00:06:14,180 --> 00:06:15,730
the U.S., in Europe, and so on.
在美国 欧洲等一些国家和地区

138
00:06:16,270 --> 00:06:18,040
They're giving more robust driving
他们提供了一些比这个方法

139
00:06:18,400 --> 00:06:19,760
controllers than this, but I
更加稳定的驾驶控制技术

140
00:06:20,080 --> 00:06:21,910
think it's still pretty remarkable and
但我认为 使用这样一个简单的

141
00:06:22,040 --> 00:06:23,190
pretty amazing how a simple
基于反向传播的神经网络

142
00:06:23,690 --> 00:06:25,440
neural network trained with back-propagation
训练出如此强大的自动驾驶汽车

143
00:06:26,140 --> 00:06:28,990
can, you know, actually learn to drive a car somewhat well.
的确是一次令人惊讶的成就