All language subtitles for 42Mark 字幕
Afrikaans
Albanian
Amharic
Arabic
Armenian
Azerbaijani
Basque
Belarusian
Bengali
Bosnian
Bulgarian
Catalan
Cebuano
Chichewa
Chinese (Simplified)
Chinese (Traditional)
Corsican
Croatian
Czech
Danish
Dutch
Esperanto
Estonian
Filipino
Finnish
French
Frisian
Galician
Georgian
German
Greek
Gujarati
Haitian Creole
Hausa
Hawaiian
Hebrew
Hindi
Hmong
Hungarian
Icelandic
Igbo
Indonesian
Irish
Italian
Japanese
Javanese
Kannada
Kazakh
Khmer
Korean
Kurdish (Kurmanji)
Kyrgyz
Lao
Latin
Latvian
Lithuanian
Luxembourgish
Macedonian
Malagasy
Malay
Malayalam
Maltese
Maori
Marathi
Mongolian
Myanmar (Burmese)
Nepali
Norwegian
Pashto
Persian
Polish
Portuguese
Punjabi
Romanian
Russian
Samoan
Scots Gaelic
Serbian
Sesotho
Shona
Sindhi
Sinhala
Slovak
Slovenian
Somali
Spanish
Sundanese
Swahili
Swedish
Tajik
Tamil
Telugu
Thai
Turkish
Ukrainian
Urdu
Uzbek
Vietnamese
Welsh
Xhosa
Yiddish
Yoruba
Zulu
Odia (Oriya)
Kinyarwanda
Turkmen
Tatar
Uyghur
Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated:
1
00:01:01,920 --> 00:01:03,320
首先先和大家说一声抱歉
2
00:01:03,320 --> 00:01:05,319
上海封控已经两个多月了
3
00:01:05,319 --> 00:01:07,319
我一直想等解封了之后
4
00:01:07,319 --> 00:01:09,639
回到办公室的棚内来给大家录制
5
00:01:09,640 --> 00:01:11,799
包括我们还有一些需要补拍的镜头
6
00:01:11,799 --> 00:01:12,840
但是两个多月了
7
00:01:12,840 --> 00:01:14,079
一直没有解封
8
00:01:14,079 --> 00:01:15,200
真的是等不住了
9
00:01:15,200 --> 00:01:17,920
所以画面跟收音可能会比较将就
10
00:01:17,920 --> 00:01:18,719
还希望大家呢
11
00:01:18,719 --> 00:01:19,519
能够多多见谅
12
00:01:19,519 --> 00:01:20,560
上期视频里面
13
00:01:20,560 --> 00:01:21,799
我们介绍了导航辅助驾驶
14
00:01:21,799 --> 00:01:24,920
其实是分为导航和辅助驾驶两个部分
15
00:01:25,159 --> 00:01:26,359
辅助驾驶可以做到
16
00:01:26,359 --> 00:01:27,640
根据前车的速度
17
00:01:27,640 --> 00:01:28,719
主动加减速
18
00:01:28,719 --> 00:01:30,679
并在车道线清晰的路段
19
00:01:30,680 --> 00:01:32,080
长时间稳定地保持
20
00:01:32,079 --> 00:01:33,560
在车道中间行驶
21
00:01:33,840 --> 00:01:35,600
在车机设置导航目的地之后
22
00:01:35,640 --> 00:01:36,400
系统则可以在
23
00:01:36,400 --> 00:01:38,160
原来辅助驾驶的基础上
24
00:01:38,359 --> 00:01:40,040
实现自动进出匝道
25
00:01:40,040 --> 00:01:41,200
自动变道超车
26
00:01:41,200 --> 00:01:42,439
自动调节限速
27
00:01:42,519 --> 00:01:44,799
上一期视频里面我们重点介绍了
28
00:01:44,799 --> 00:01:47,119
特理蔚鹏实测的匝道表现
29
00:01:47,120 --> 00:01:48,040
如果你还没有看过
30
00:01:48,040 --> 00:01:48,920
我强烈推荐你
31
00:01:48,920 --> 00:01:50,359
先回去看上一期
32
00:01:50,359 --> 00:01:51,040
本期节目呢
33
00:01:51,040 --> 00:01:53,240
我们会对比这 4 辆车的变道策略
34
00:01:53,239 --> 00:01:54,439
和人机共驾策略
35
00:01:54,439 --> 00:01:56,200
并给出最终评分
36
00:02:00,239 --> 00:02:02,280
匝道能力是导航辅助驾驶的基础
37
00:02:02,319 --> 00:02:04,839
但是真正决定效率和体验的
38
00:02:04,879 --> 00:02:07,479
是被很多人忽视的变道策略
39
00:02:07,719 --> 00:02:09,080
毕竟在一段行程中
40
00:02:09,080 --> 00:02:11,840
匝道占比一定是远低于变道的
41
00:02:12,000 --> 00:02:14,639
在江浙沪车流相对密集的高速环境下
42
00:02:14,639 --> 00:02:17,079
我们很难一直在一条车道上行驶
43
00:02:17,080 --> 00:02:20,200
我相信在超车道龟速行驶的情况
44
00:02:20,199 --> 00:02:21,919
大家肯定都遇到过
45
00:02:21,919 --> 00:02:24,159
所以一个出色的变道策略
46
00:02:24,159 --> 00:02:25,439
就很关键了
47
00:02:25,439 --> 00:02:26,560
为了评价出这 4 辆车
48
00:02:26,560 --> 00:02:27,879
变道策略的好坏
49
00:02:27,879 --> 00:02:29,359
我们从变道速度差
50
00:02:29,400 --> 00:02:30,960
和目标车道空间要求
51
00:02:31,199 --> 00:02:31,839
2个维度
52
00:02:31,879 --> 00:02:33,240
设置了 6 个测试项目
53
00:02:33,800 --> 00:02:34,960
我们在跑高速的时候
54
00:02:35,000 --> 00:02:36,240
萌生变道想法
55
00:02:36,280 --> 00:02:38,719
最主要的原因就是前车慢了
56
00:02:39,000 --> 00:02:40,680
所以和前车的速度差
57
00:02:40,719 --> 00:02:42,919
就是触发变道的第一道门槛
58
00:02:43,319 --> 00:02:45,079
大家也可以在弹幕里分享一下
59
00:02:45,120 --> 00:02:46,680
前车一般慢到什么程度
60
00:02:46,719 --> 00:02:48,359
你会受不了主动变道
61
00:02:48,759 --> 00:02:50,120
首先我们得明确一点
62
00:02:50,159 --> 00:02:51,199
在不同的车道
63
00:02:51,240 --> 00:02:53,560
大家变道的积极性是不一样的
64
00:02:54,039 --> 00:02:55,599
大部分人在上了高速之后
65
00:02:55,800 --> 00:02:57,760
都会第一时间变到最左侧车道
66
00:02:57,800 --> 00:02:59,560
然后进入一个巡航的状态
67
00:02:59,879 --> 00:03:01,240
即使前面有慢车
68
00:03:01,280 --> 00:03:03,240
只要不是慢得特别过分的
69
00:03:03,280 --> 00:03:04,520
都不会主动变道
70
00:03:04,840 --> 00:03:05,680
而在右侧车道
71
00:03:05,719 --> 00:03:06,319
大部分人呢
72
00:03:06,360 --> 00:03:08,320
会更加积极地往左侧车道变
73
00:03:08,360 --> 00:03:10,400
毕竟左侧车道大部分情况下
74
00:03:10,439 --> 00:03:12,120
都会比右侧更加通畅
75
00:03:12,520 --> 00:03:13,120
说白了
76
00:03:13,159 --> 00:03:14,799
就是该积极的时候得积极
77
00:03:14,800 --> 00:03:16,439
该克制的时候得克制
78
00:03:16,919 --> 00:03:17,839
我们实际体验下来
79
00:03:17,879 --> 00:03:19,479
理想 ONE 的标定最细腻
80
00:03:19,520 --> 00:03:22,680
根据环境和变道方向都做了区分
81
00:03:22,879 --> 00:03:23,919
在三车道的环境下
82
00:03:23,960 --> 00:03:26,040
触发从左往右变道的速度差
83
00:03:26,159 --> 00:03:27,280
是 15 km/h
84
00:03:27,319 --> 00:03:29,599
也就是说限速 120 的车道下
85
00:03:29,639 --> 00:03:31,519
前车的速度低于 105 km/h
86
00:03:31,520 --> 00:03:32,960
系统就会触发变道
87
00:03:35,599 --> 00:03:37,079
变了
88
00:03:37,120 --> 00:03:39,319
这个速度差既不会因为太小
89
00:03:39,360 --> 00:03:40,760
导致变道过于频繁
90
00:03:40,800 --> 00:03:41,960
也不会因为过大
91
00:03:42,000 --> 00:03:43,879
让驾驶员觉得很拖沓
92
00:03:44,199 --> 00:03:45,399
这里需要注意一点是
93
00:03:45,439 --> 00:03:46,840
这个速度差并不需要
94
00:03:46,879 --> 00:03:49,079
真的等车速降到 105 了
95
00:03:49,120 --> 00:03:49,879
才会变道
96
00:03:50,120 --> 00:03:51,840
只要你的车速跟前车的速度差
97
00:03:51,879 --> 00:03:53,359
已经达到了 15
98
00:03:53,560 --> 00:03:54,800
那它就会主动变道
99
00:03:54,840 --> 00:03:56,240
从右往左变道的速度差
100
00:03:56,360 --> 00:03:57,760
则在 5—10 km/h
101
00:03:57,800 --> 00:03:59,480
考虑到左侧车道的速度
102
00:03:59,599 --> 00:04:01,039
一定是快于右侧的
103
00:04:01,280 --> 00:04:02,400
这样的标定逻辑
104
00:04:02,479 --> 00:04:03,840
可以更加积极主动地
105
00:04:03,879 --> 00:04:05,439
往左侧汇入快车道
106
00:04:05,479 --> 00:04:06,840
你看它在这个车道
107
00:04:06,840 --> 00:04:08,680
可以非常积极地向左边变
108
00:04:11,319 --> 00:04:13,719
你看它现在又好像激进了一点
109
00:04:14,000 --> 00:04:15,199
在两车道的环境下
110
00:04:15,199 --> 00:04:17,279
触发从左往右变道的速度差
111
00:04:17,399 --> 00:04:19,319
是 15—20 km/h 左右
112
00:04:19,639 --> 00:04:20,879
从右往左变的速度差
113
00:04:20,920 --> 00:04:22,560
依然是 5—10 km/h
114
00:04:22,600 --> 00:04:24,280
这样的策略可以抑制系统
115
00:04:24,319 --> 00:04:26,199
变到右侧货车道的积极性
116
00:04:26,240 --> 00:04:28,160
同时也提高了从慢车道
117
00:04:28,199 --> 00:04:29,879
变回到快车道的积极性
118
00:04:32,879 --> 00:04:35,360
这次速度差只有 20 不到
119
00:04:35,560 --> 00:04:37,199
小鹏和蔚来的策略类似
120
00:04:37,240 --> 00:04:38,480
在三车道的情况下
121
00:04:38,600 --> 00:04:40,680
左右变道的速度差没有区别
122
00:04:41,399 --> 00:04:42,319
非常聪明
123
00:04:42,319 --> 00:04:43,680
把前车给它超了
124
00:04:45,480 --> 00:04:46,960
刚才速度差是 10 左右
125
00:04:47,240 --> 00:04:48,199
两车道的情况下
126
00:04:48,360 --> 00:04:50,280
左变右速度差会略高一些
127
00:04:53,839 --> 00:04:54,679
舒服啊
128
00:04:54,720 --> 00:04:56,280
它感觉但是这么看下来
129
00:04:56,279 --> 00:04:57,439
感觉它在两车道
130
00:04:57,480 --> 00:04:59,800
跟四车道下都差不多
131
00:05:00,000 --> 00:05:02,160
都是压速度压到 15 左右
132
00:05:02,199 --> 00:05:04,560
特斯拉的策略是最简单粗暴的
133
00:05:04,600 --> 00:05:07,080
只要车速差达到 15 km/h
134
00:05:07,120 --> 00:05:08,120
会立马变道
135
00:05:08,120 --> 00:05:09,280
不论什么环境
136
00:05:09,319 --> 00:05:11,839
现在速度差是 10 码左右
137
00:05:11,839 --> 00:05:13,199
还没有触发变道
138
00:05:13,240 --> 00:05:15,639
110 106 105
139
00:05:15,680 --> 00:05:16,400
还没有触发变道
140
00:05:16,399 --> 00:05:17,239
触发变道了
141
00:05:17,560 --> 00:05:19,040
看到这里你是不是有疑问了
142
00:05:19,079 --> 00:05:20,120
为什么这个数据
143
00:05:20,160 --> 00:05:22,200
和实际体验的区别这么大
144
00:05:22,560 --> 00:05:23,839
为什么感觉理想的体验
145
00:05:24,000 --> 00:05:24,759
没有什么亮点
146
00:05:24,800 --> 00:05:26,800
反而特斯拉是最频繁的
147
00:05:27,279 --> 00:05:29,239
核心原因就是对目标车道
148
00:05:29,279 --> 00:05:30,639
空间要求的不同
149
00:05:31,319 --> 00:05:32,120
上面一项测试
150
00:05:32,199 --> 00:05:34,599
我们明确了不同车道慢的标准
151
00:05:34,639 --> 00:05:36,639
但是真正决定要不要变道的
152
00:05:36,839 --> 00:05:38,439
是目标车道的空间
153
00:05:38,800 --> 00:05:39,360
说白了
154
00:05:39,399 --> 00:05:41,319
就是有没有位置让你变过去
155
00:05:41,360 --> 00:05:42,800
这是第二道门槛
156
00:05:43,319 --> 00:05:45,159
回想一下我们自己在开车的时候
157
00:05:45,199 --> 00:05:47,759
变道前不仅会看变道的那条车道
158
00:05:47,800 --> 00:05:49,000
前面有没有车
159
00:05:49,160 --> 00:05:50,480
位置会不会影响到我们变道
160
00:05:50,519 --> 00:05:52,240
也会观察后面有没有车
161
00:05:52,279 --> 00:05:53,759
位置会不会影响到我们变道
162
00:05:53,800 --> 00:05:55,000
然后综合判断
163
00:05:55,000 --> 00:05:56,439
最终能否变道
164
00:05:57,079 --> 00:05:58,479
根据目标车道位置不同
165
00:05:58,519 --> 00:06:00,039
我们设置了 5 个环境
166
00:06:01,680 --> 00:06:03,160
这里特斯拉对目标车道
167
00:06:03,199 --> 00:06:04,479
空间要求是最小的
168
00:06:04,519 --> 00:06:06,560
前方车辆只要快于我方车辆
169
00:06:06,600 --> 00:06:08,400
车辆位置处于负值
170
00:06:08,399 --> 00:06:09,399
即可发起变道
171
00:06:09,560 --> 00:06:11,360
后方有车的话也只是空间
172
00:06:11,399 --> 00:06:13,799
只要足够就会立刻发起变道
173
00:06:17,040 --> 00:06:19,040
完全不顾后面汉的死活
174
00:06:19,079 --> 00:06:20,279
直接一把拉了过来
175
00:06:20,319 --> 00:06:21,680
不过汉没有闪远光灯
176
00:06:21,959 --> 00:06:22,839
非常地给力
177
00:06:23,800 --> 00:06:25,000
识别到了
178
00:06:25,040 --> 00:06:26,160
又要开始了
179
00:06:26,839 --> 00:06:28,839
特斯拉直接别别人
180
00:06:28,879 --> 00:06:30,399
特斯拉 别人按喇叭也不管
181
00:06:30,879 --> 00:06:33,120
这样的策略和激进的人类驾驶员
182
00:06:33,279 --> 00:06:34,439
非常相似
183
00:06:34,720 --> 00:06:35,880
也会为自己争取到
184
00:06:35,920 --> 00:06:37,199
最大的变道空间
185
00:06:37,240 --> 00:06:38,759
在车流密集的环境下
186
00:06:38,879 --> 00:06:40,199
仍然可以自由地变道
187
00:06:40,480 --> 00:06:41,960
这背后也体现出了
188
00:06:41,959 --> 00:06:43,879
特斯拉扎实的变道基础
189
00:06:45,240 --> 00:06:46,600
这时候我们
190
00:06:50,040 --> 00:06:51,400
先打灯变道
191
00:06:53,160 --> 00:06:54,720
我们把后面的奔驰 C 别住了
192
00:06:55,519 --> 00:06:56,799
但是遗憾就是特斯拉呢
193
00:06:56,920 --> 00:06:57,920
真的不够懂
194
00:06:57,959 --> 00:06:59,079
中国道路的特点
195
00:06:59,120 --> 00:06:59,879
变道的速度差
196
00:06:59,879 --> 00:07:01,319
标定非常粗糙
197
00:07:01,560 --> 00:07:02,879
所以在两车道的环境下
198
00:07:02,920 --> 00:07:05,680
特斯拉会特别积极地往右变道
199
00:07:05,959 --> 00:07:07,159
但是受限于感知盲区
200
00:07:07,399 --> 00:07:09,799
经常是变过去了才发现有大车
201
00:07:09,839 --> 00:07:10,919
然后速度非常慢
202
00:07:10,959 --> 00:07:12,359
然后又不得不再变回来
203
00:07:12,399 --> 00:07:14,159
出现了多次的无效变道
204
00:07:14,360 --> 00:07:15,600
体验并不算好
205
00:07:15,920 --> 00:07:16,680
对空间要求
206
00:07:16,680 --> 00:07:18,000
第二低的是蔚来
207
00:07:18,040 --> 00:07:20,160
怎么样是不是有点出乎意料
208
00:07:20,199 --> 00:07:21,439
同样是前方的车辆
209
00:07:21,480 --> 00:07:22,680
只要快于我方车辆
210
00:07:22,720 --> 00:07:24,360
车辆的位置处于负值
211
00:07:24,399 --> 00:07:25,359
即可发起变道
212
00:07:25,639 --> 00:07:26,879
每次在这种情况下
213
00:07:26,920 --> 00:07:28,000
触发变道的感觉
214
00:07:28,160 --> 00:07:29,320
真的很惊艳
215
00:07:31,079 --> 00:07:33,839
变 终于舒服了
216
00:07:35,879 --> 00:07:37,199
对后方车辆的距离要求
217
00:07:37,319 --> 00:07:38,920
也只是略高于特斯拉
218
00:07:38,959 --> 00:07:40,039
但是也非常接近
219
00:07:40,079 --> 00:07:41,439
人类驾驶员的判断
220
00:07:42,879 --> 00:07:43,719
来了
221
00:07:44,560 --> 00:07:46,360
现在这个车后面离我们近
222
00:07:47,839 --> 00:07:49,279
这个能变过来
223
00:07:50,519 --> 00:07:51,120
特别好
224
00:07:51,600 --> 00:07:52,760
后面还有个特斯拉
225
00:07:52,759 --> 00:07:53,599
有点紧张
226
00:07:53,639 --> 00:07:55,000
就不变道了吧
227
00:08:01,680 --> 00:08:03,680
往右超车了
228
00:08:03,720 --> 00:08:06,240
聪明
229
00:08:06,279 --> 00:08:08,039
蔚来能把毫米波雷达
230
00:08:08,079 --> 00:08:09,359
调校到这样的水平
231
00:08:09,399 --> 00:08:11,199
完全是我们意料之外的
232
00:08:11,720 --> 00:08:13,240
这都能变过去
233
00:08:14,199 --> 00:08:15,719
牛啊
234
00:08:16,879 --> 00:08:18,480
我觉得蔚来
235
00:08:18,480 --> 00:08:21,040
真的是把毫米波用到了极限
236
00:08:21,600 --> 00:08:23,000
结合变道速度差的标定
237
00:08:23,040 --> 00:08:25,040
蔚来的表现和驾驶员的预期
238
00:08:25,240 --> 00:08:26,199
都非常接近
239
00:08:26,600 --> 00:08:27,480
大部分情况下
240
00:08:27,480 --> 00:08:28,680
驾驶员想变了
241
00:08:28,680 --> 00:08:29,480
15秒之内
242
00:08:29,519 --> 00:08:31,000
车都会发起变道
243
00:08:31,360 --> 00:08:32,720
兼顾了效率和体验
244
00:08:32,759 --> 00:08:34,639
是这 4 辆车中表现最好的
245
00:08:34,919 --> 00:08:37,599
小鹏 P7 对前车位置的要求略高
246
00:08:37,639 --> 00:08:39,720
在车速快于我方车辆的情况下
247
00:08:39,759 --> 00:08:41,799
需要前车的车尾平齐
248
00:08:42,080 --> 00:08:43,240
才会发起变道
249
00:08:43,480 --> 00:08:45,000
这个也还算合理
250
00:08:45,039 --> 00:08:46,559
但是对后车的要求
251
00:08:46,600 --> 00:08:48,080
可以说就是非常高了
252
00:08:48,120 --> 00:08:49,720
不仅需要一个较大的空间
253
00:08:49,759 --> 00:08:50,879
才能变道过去
254
00:08:50,919 --> 00:08:53,360
如果后车有轻微加速的趋势
255
00:08:53,639 --> 00:08:55,399
系统也不会发起变道
256
00:08:55,399 --> 00:08:56,919
而且我发现一个问题
257
00:08:56,919 --> 00:09:00,599
它一旦等待后车来的时候
258
00:09:00,600 --> 00:09:01,840
它就取消变道
259
00:09:01,840 --> 00:09:03,600
之后它这个变道策略
260
00:09:03,600 --> 00:09:04,759
将会变得非常的保守
261
00:09:04,759 --> 00:09:07,639
其实左边现在的车离我们非常远
262
00:09:07,639 --> 00:09:09,720
但是它也没有选择变道
263
00:09:09,720 --> 00:09:11,720
所以在车流量相对较大的环境下
264
00:09:11,919 --> 00:09:13,879
经常卡死在一条慢车道
265
00:09:14,679 --> 00:09:19,599
空了
266
00:09:19,600 --> 00:09:20,560
然后变了
267
00:09:20,679 --> 00:09:21,559
经过我们统计
268
00:09:21,600 --> 00:09:23,879
小鹏最长一次被卡在慢车道
269
00:09:23,919 --> 00:09:24,959
接近 5 分钟
270
00:09:25,000 --> 00:09:26,759
被 31 辆车超过
271
00:09:27,000 --> 00:09:28,200
真的是非常的崩溃
272
00:09:35,320 --> 00:09:36,480
小鹏现在开得慢了
273
00:09:36,480 --> 00:09:37,320
遇到慢车了
274
00:09:41,080 --> 00:09:42,440
它后方有快车经过
275
00:09:42,519 --> 00:09:44,679
它一分钟都没有变道
276
00:09:44,679 --> 00:09:45,839
现在已经导致我
277
00:09:45,879 --> 00:09:47,080
一直在被别人超了
278
00:09:47,080 --> 00:09:48,840
很奇怪
279
00:09:52,480 --> 00:09:55,800
不行啊 一直不让我变道了
280
00:09:56,440 --> 00:09:57,760
我现在开得很慢才 80
281
00:10:03,399 --> 00:10:04,519
对目标车道空间
282
00:10:04,519 --> 00:10:06,120
要求最高的是理想
283
00:10:06,480 --> 00:10:07,600
在目标车道前车
284
00:10:07,600 --> 00:10:09,000
快于我方车辆的前提下
285
00:10:09,240 --> 00:10:10,720
车辆的位置处于正值
286
00:10:10,759 --> 00:10:11,759
才会发起变道
287
00:10:12,039 --> 00:10:13,719
对后方车辆距离的要求
288
00:10:13,879 --> 00:10:15,240
和小鹏是差不多的
289
00:10:20,320 --> 00:10:21,480
又变回去了
290
00:10:22,679 --> 00:10:25,079
所以再出色的变道速度差标定
291
00:10:25,120 --> 00:10:27,360
面对极高的空间要求
292
00:10:27,399 --> 00:10:28,639
也就不灵了
293
00:10:28,679 --> 00:10:30,319
对于急性子测试员包哥来说
294
00:10:30,480 --> 00:10:32,480
可以说是非常崩溃了
295
00:10:32,519 --> 00:10:34,480
差不多该变了吧
296
00:10:36,039 --> 00:10:37,799
咱说真的这都 90 了
297
00:10:37,799 --> 00:10:39,319
咱还不变
298
00:10:39,960 --> 00:10:40,920
咱这就过分了
299
00:10:40,919 --> 00:10:42,079
咱这
300
00:10:43,480 --> 00:10:45,519
咱这真的过分了
301
00:10:45,519 --> 00:10:46,439
真的
302
00:10:46,679 --> 00:10:48,000
不是我跟你说 80 了
303
00:10:50,080 --> 00:10:51,680
80了你不变
304
00:10:52,519 --> 00:10:53,879
真的 80 了
305
00:10:53,919 --> 00:10:55,599
快车道 80 了大哥
306
00:10:56,559 --> 00:10:57,919
咱变一变吧
307
00:10:57,919 --> 00:10:59,120
变了
308
00:10:59,799 --> 00:11:00,519
这我得劝它
309
00:11:00,519 --> 00:11:01,480
它才变
310
00:11:01,480 --> 00:11:03,200
上面两项测试的参数设定
311
00:11:03,320 --> 00:11:05,040
我们都希望一辆车的变道
312
00:11:05,159 --> 00:11:06,799
能够更加主动一些
313
00:11:06,840 --> 00:11:08,360
但是变得越多
314
00:11:08,399 --> 00:11:10,279
出错的可能性也就越大
315
00:11:10,639 --> 00:11:12,480
为了制衡出现的无效变道
316
00:11:12,480 --> 00:11:13,399
我们还统计了
317
00:11:13,440 --> 00:11:15,040
百公里无效变道次数
318
00:11:15,320 --> 00:11:16,040
不出意外的是
319
00:11:16,039 --> 00:11:17,679
特斯拉的无效变道次数最多
320
00:11:17,720 --> 00:11:19,759
百公里接近 7.5 次
321
00:11:19,840 --> 00:11:21,920
蔚来第二只有 3.7 次
322
00:11:22,120 --> 00:11:23,639
小鹏第三 3.1 次
323
00:11:23,679 --> 00:11:26,039
理想最少不到 2.5 次
324
00:11:26,399 --> 00:11:27,879
特斯拉大部分无效变道
325
00:11:28,080 --> 00:11:29,920
都主要集中在变道过于积极
326
00:11:29,960 --> 00:11:31,240
但是感知能力受限
327
00:11:31,279 --> 00:11:33,679
探测不到目标车道远处的慢车
328
00:11:33,720 --> 00:11:34,519
变道之后
329
00:11:34,519 --> 00:11:35,960
反而被慢车压了速度
330
00:11:36,279 --> 00:11:37,759
然后丢了原来的位置
331
00:11:38,480 --> 00:11:39,680
疯狂变道
332
00:11:39,679 --> 00:11:40,639
检测到了
333
00:11:42,879 --> 00:11:44,000
特斯拉太活泼了
334
00:11:44,000 --> 00:11:44,879
又要往右变
335
00:11:48,639 --> 00:11:50,480
变了之后就 say byebye
336
00:11:54,120 --> 00:11:55,080
886
337
00:11:55,080 --> 00:11:56,960
886特斯拉
338
00:11:56,960 --> 00:11:57,920
886
339
00:11:58,200 --> 00:11:59,720
特斯拉一直在无效变道
340
00:12:01,279 --> 00:12:02,159
它又变回来了
341
00:12:03,480 --> 00:12:05,080
它全都是无效变道特斯拉
342
00:12:05,600 --> 00:12:06,240
这样的问题
343
00:12:06,279 --> 00:12:08,199
在两车道环境下尤为明显
344
00:12:08,919 --> 00:12:10,679
前面 要变道了
345
00:12:10,759 --> 00:12:12,360
速度差差 15
346
00:12:12,519 --> 00:12:13,639
然后变完道之后
347
00:12:13,679 --> 00:12:15,079
发现前面有大车
348
00:12:15,120 --> 00:12:17,240
好 无效变道完美
349
00:12:17,799 --> 00:12:19,359
而理想 小鹏的无效变道
350
00:12:19,399 --> 00:12:20,120
主要集中在
351
00:12:20,120 --> 00:12:22,480
对目标车道的空间要求过高
352
00:12:22,759 --> 00:12:25,080
超车之后没法及时回到快车道
353
00:12:25,559 --> 00:12:27,319
所以没什么遮挡
354
00:12:28,360 --> 00:12:30,960
104对
355
00:12:30,960 --> 00:12:34,160
这次是差了 16
356
00:12:38,120 --> 00:12:40,879
但这次
357
00:12:40,879 --> 00:12:41,799
怎么说呢
358
00:12:41,799 --> 00:12:43,719
你觉得这种能算有效（变道）吗
359
00:12:47,200 --> 00:12:47,759
因为它
360
00:12:47,759 --> 00:12:48,879
其实你看前面
361
00:12:48,879 --> 00:12:50,159
当然是无效的
362
00:12:50,159 --> 00:12:50,559
没有用
363
00:12:50,559 --> 00:12:51,759
你看它一会变回来了
364
00:12:52,440 --> 00:12:53,880
在车流量较小的环境下
365
00:12:54,000 --> 00:12:56,320
这 4 辆车的表现都非常出色
366
00:12:56,679 --> 00:12:57,319
预判变道
367
00:12:57,320 --> 00:12:58,160
行云流水
368
00:13:00,759 --> 00:13:01,519
要超车了
369
00:13:01,519 --> 00:13:02,759
现在小鹏开始牛了 要超车了
370
00:13:05,440 --> 00:13:06,440
116
371
00:13:06,440 --> 00:13:07,960
5 km/h就超了
372
00:13:11,120 --> 00:13:12,120
闪的
373
-->
你看一下
374
-->
又超车了
375
00:13:15,600 --> 00:13:17,040
预判变道
376
00:13:17,360 --> 00:13:18,600
它虽然速度没有被压
377
00:13:18,600 --> 00:13:20,320
但是它可能已经探测到前面
378
00:13:20,519 --> 00:13:22,279
这个车的速度是比我慢的
379
00:13:22,320 --> 00:13:23,960
然后我就提前先变过来
380
00:13:23,960 --> 00:13:25,480
在不影响我自己车速的情况下
381
00:13:25,480 --> 00:13:26,399
把它超了
382
00:13:26,399 --> 00:13:28,039
但是车流量一旦变大
383
00:13:28,039 --> 00:13:29,759
则暴露了很多问题
384
00:13:35,639 --> 00:13:37,399
每次老有车打断它
385
00:13:37,399 --> 00:13:38,639
蔚来的综合体验最佳
386
00:13:38,679 --> 00:13:39,639
剩下的三位
387
00:13:39,679 --> 00:13:42,279
分数接近属于一个水平
388
00:13:42,320 --> 00:13:43,800
特斯拉没有根据车道
389
00:13:43,799 --> 00:13:44,959
区分变道策略
390
00:13:44,960 --> 00:13:47,000
而小鹏 理想则对目标车道的
391
00:13:47,000 --> 00:13:48,759
空间要求过高
392
00:13:48,759 --> 00:13:49,480
这里多说一点
393
00:13:49,519 --> 00:13:51,319
2.7版本下的小鹏 P7
394
00:13:51,360 --> 00:13:52,440
在空旷路况下
395
00:13:52,480 --> 00:13:54,480
变道时机的选择和判断
396
00:13:54,639 --> 00:13:56,439
准确性都有一定的提高
397
00:13:56,600 --> 00:13:58,360
但是在车流量较大的环境下
398
00:13:58,600 --> 00:13:59,279
变道积极性
399
00:13:59,279 --> 00:14:00,879
明显低于 2.6 的版本
400
00:14:01,240 --> 00:14:02,600
有同样感受的鹏友
401
00:14:02,639 --> 00:14:04,399
也可以在弹幕里回应一下
402
00:14:07,559 --> 00:14:08,799
牛
403
00:14:08,799 --> 00:14:09,879
这个牛 2 km
404
00:14:10,759 --> 00:14:12,439
直接识别到前面慢车直接变
405
00:14:15,799 --> 00:14:17,319
这个就显得很奇怪
406
00:14:17,320 --> 00:14:18,879
这个小鹏让人搞不懂
407
00:14:18,879 --> 00:14:20,439
之前慢得要死
408
00:14:20,440 --> 00:14:22,320
降到 90 多它都不变
409
00:14:22,320 --> 00:14:24,600
这次 2 km 它就变回来了
410
00:14:24,600 --> 00:14:26,080
这个路况好和路况差
411
00:14:26,080 --> 00:14:28,600
这个小鹏 NGP 的差别很大
412
00:14:28,600 --> 00:14:30,360
我也向小鹏的朋友求证了一下
413
00:14:30,360 --> 00:14:31,919
因为他们在调研里发现
414
00:14:31,919 --> 00:14:33,639
比较多的用户反馈
415
00:14:33,639 --> 00:14:35,080
变道过于积极
416
00:14:35,120 --> 00:14:36,279
所以降低了积极性
417
00:14:38,720 --> 00:14:40,840
90 km/h
418
00:14:40,840 --> 00:14:42,600
（速度差）30 km/h
419
00:14:42,600 --> 00:14:43,680
在这次测试里
420
00:14:43,720 --> 00:14:44,440
我们也发现
421
00:14:44,480 --> 00:14:45,720
不同的驾驶员性格
422
00:14:45,720 --> 00:14:48,279
对变道积极性的要求也有差异
423
00:14:48,519 --> 00:14:49,720
所以如何平衡
424
00:14:49,759 --> 00:14:51,200
积极和保守
425
00:14:51,240 --> 00:14:52,440
是高阶辅助驾驶
426
00:14:52,480 --> 00:14:53,720
带来的新问题
427
00:14:56,240 --> 00:14:56,919
特殊场景里
428
00:14:56,919 --> 00:14:58,399
我们设置了雨天路况
429
00:14:58,399 --> 00:14:59,199
施工路况
430
00:14:59,200 --> 00:15:01,520
高精地图与实际道路情况不符
431
00:15:01,519 --> 00:15:03,079
三种使用过程中
432
00:15:03,120 --> 00:15:05,120
可能会遇到的非常规路段
433
00:15:05,480 --> 00:15:06,480
首先我们还是建议
434
00:15:06,600 --> 00:15:07,680
碰到这三种情况
435
00:15:07,720 --> 00:15:09,480
大家一定要第一时间接管
436
00:15:09,519 --> 00:15:10,360
确保安全
437
00:15:10,639 --> 00:15:12,519
我们的尝试也是为了帮助大家
438
00:15:12,559 --> 00:15:14,639
摸索出系统能力的边界
439
00:15:15,120 --> 00:15:15,840
雨天路况下
440
00:15:15,879 --> 00:15:17,679
表现最好的是小鹏和理想
441
00:15:17,720 --> 00:15:18,960
只有极少数的场景
442
00:15:19,000 --> 00:15:21,200
会出现辅助驾驶降级的场景
443
00:15:21,600 --> 00:15:22,800
毕竟在 IMU 惯导
444
00:15:22,840 --> 00:15:24,280
和高精定位的加持下
445
00:15:24,559 --> 00:15:26,479
蒙上眼睛车都知道自己在哪
446
00:15:27,240 --> 00:15:28,120
其次是特斯拉
447
00:15:28,159 --> 00:15:29,360
只要雨量达到中雨
448
00:15:29,399 --> 00:15:30,439
特斯拉的导航辅助
449
00:15:30,559 --> 00:15:33,079
就会主动降低为基础辅助驾驶
450
00:15:33,279 --> 00:15:34,319
但是好在
451
00:15:34,360 --> 00:15:36,720
基础辅助驾驶的表现十分稳定
452
00:15:36,720 --> 00:15:39,000
而且打灯变道也可以正常使用
453
00:15:39,279 --> 00:15:40,360
最后是蔚来
454
00:15:40,399 --> 00:15:41,679
同样是中雨的情况
455
00:15:41,679 --> 00:15:44,719
系统则存在完全无法开启的可能
456
00:15:47,279 --> 00:15:48,480
NIO Pilot未启动
457
00:15:48,480 --> 00:15:50,399
请稍后再试吧
458
00:15:50,399 --> 00:15:51,799
高速上大部分施工
459
00:15:52,200 --> 00:15:53,400
都会非常规矩地
460
00:15:53,440 --> 00:15:55,160
用锥筒摆出一道斜线
461
00:15:55,200 --> 00:15:56,200
把施工的车道
462
00:15:56,320 --> 00:15:58,200
引导到旁边空旷的车道上
463
00:15:58,639 --> 00:16:00,399
恰巧我们这次测试的 4 辆车
464
00:16:00,559 --> 00:16:02,279
都具备锥筒识别能力
465
00:16:02,679 --> 00:16:03,839
但是高速行驶中
466
00:16:03,879 --> 00:16:05,000
想要识别到锥筒
467
00:16:05,039 --> 00:16:06,679
这样较小的物体
468
00:16:06,919 --> 00:16:08,679
对感知算法的要求极高
469
00:16:09,159 --> 00:16:10,839
表现最好的还是特斯拉
470
00:16:10,879 --> 00:16:12,240
锥筒位置识别稳定
471
00:16:12,279 --> 00:16:14,639
而且系统对锥筒十分敏感
472
00:16:14,679 --> 00:16:16,279
只要路边上出现锥筒
473
00:16:16,600 --> 00:16:17,480
就会触发变道
474
00:16:17,519 --> 00:16:18,840
结合极强的变道能力
475
00:16:19,399 --> 00:16:20,319
大部分场景下
476
00:16:20,320 --> 00:16:22,120
都能提前远离锥筒
477
00:16:22,960 --> 00:16:25,960
特斯拉自动调成了 110
478
00:16:31,240 --> 00:16:32,240
小鹏和理想 蔚来
479
00:16:32,440 --> 00:16:33,920
虽然也能识别到锥筒
480
00:16:33,960 --> 00:16:35,320
但是受限于感知能力
481
00:16:35,360 --> 00:16:36,440
探测到锥筒的时候
482
00:16:36,559 --> 00:16:38,319
距离可能已经比较近了
483
00:16:38,600 --> 00:16:40,279
所以留给系统变道的时间
484
00:16:40,320 --> 00:16:41,160
非常有限
485
00:16:41,399 --> 00:16:42,480
再加上我们刚刚说过了
486
00:16:42,519 --> 00:16:44,519
对目标车道的空间要求比较高
487
00:16:44,559 --> 00:16:47,479
这也导致避让锥筒的成功率比较低
488
00:16:47,480 --> 00:16:48,680
高精地图与实际道路
489
00:16:48,720 --> 00:16:49,639
不符的处理方式
490
00:16:49,799 --> 00:16:51,519
是针对高精地图车型
491
00:16:51,559 --> 00:16:52,799
特设的测试项目
492
00:16:52,840 --> 00:16:54,120
我们先来看视频
493
00:16:57,240 --> 00:16:58,240
撞上去了
494
00:16:58,240 --> 00:16:59,440
这个真不敢了
495
00:16:59,440 --> 00:17:01,880
我一直 这真不行
496
00:17:02,759 --> 00:17:04,000
来了
497
00:17:04,039 --> 00:17:05,319
怎么回事
498
00:17:05,319 --> 00:17:07,879
汇不汇
499
00:17:08,960 --> 00:17:10,840
冲出去了
500
00:17:10,839 --> 00:17:13,199
它完全相信高精地图啊
501
00:17:13,200 --> 00:17:14,400
直接出去了
502
00:17:15,200 --> 00:17:17,519
首先在面对施工改道路段的时候
503
00:17:17,519 --> 00:17:20,079
我们强烈推荐大家直接接管
504
00:17:20,079 --> 00:17:21,359
不要做危险尝试
505
00:17:21,680 --> 00:17:22,640
但是通过这项测试
506
00:17:22,799 --> 00:17:24,159
仍然可以很明显地发现
507
00:17:24,359 --> 00:17:25,719
在施工改道路段
508
00:17:25,759 --> 00:17:27,879
即使标注了新的车道线
509
00:17:28,160 --> 00:17:29,440
小鹏和理想仍然会
510
00:17:29,480 --> 00:17:31,400
按照高精地图的路线行驶
511
00:17:31,880 --> 00:17:33,640
也就是说在出现实际道路
512
00:17:33,680 --> 00:17:35,360
与高精地图不符的时候
513
00:17:35,400 --> 00:17:37,200
系统更愿意相信
514
00:17:37,240 --> 00:17:38,559
高精地图的数据
515
00:17:38,559 --> 00:17:40,440
视野可能会被大车遮挡
516
00:17:44,839 --> 00:17:46,919
退了 NPG 要退了
517
00:17:47,119 --> 00:17:49,039
NGP退了
518
00:17:49,039 --> 00:17:50,200
改道它退了
519
00:17:50,200 --> 00:17:51,559
其实有个小问题
520
00:17:51,599 --> 00:17:52,519
为什么它会这样
521
00:17:52,720 --> 00:17:54,160
你看其实这个路上的
522
00:17:54,200 --> 00:17:55,400
车道线是很明显
523
00:17:55,440 --> 00:17:57,200
但是它完全是按照高精地图一样
524
00:17:57,240 --> 00:17:58,680
咣一下往前冲着走
525
00:18:01,839 --> 00:18:02,759
很奇怪
526
00:18:02,799 --> 00:18:04,519
说明它视觉在里面没发挥作用
527
00:18:04,559 --> 00:18:07,119
说明那它整一个 NGP
528
00:18:07,160 --> 00:18:08,720
几乎完全是靠着
529
00:18:08,759 --> 00:18:10,359
高精度地图数据
530
00:18:10,359 --> 00:18:12,199
天眼也有不灵的时候
531
00:18:12,799 --> 00:18:13,919
很尴尬
532
00:18:13,960 --> 00:18:15,400
对我来说这个落差感挺大的
533
00:18:15,440 --> 00:18:16,400
因为有高精地图
534
00:18:16,400 --> 00:18:16,960
有 NGP 的时候
535
00:18:16,960 --> 00:18:19,640
其实这个车按导航辅助驾驶
536
00:18:19,640 --> 00:18:20,440
非常的灵
537
00:18:20,480 --> 00:18:22,319
但一旦高精地图失效之后
538
00:18:22,319 --> 00:18:24,000
明显地感觉这个等级
539
00:18:24,000 --> 00:18:25,440
一下子下了一个大级
540
00:18:25,480 --> 00:18:26,799
就是体验明显打折
541
00:18:26,799 --> 00:18:27,759
就有点难受
542
00:18:28,039 --> 00:18:29,559
同样有高精地图辅助的蔚来
543
00:18:29,759 --> 00:18:31,920
则是视觉优先于高精地图
544
00:18:31,960 --> 00:18:33,360
这一点绝对好评
545
00:18:33,400 --> 00:18:35,040
希望大家向蔚来学习
546
00:18:37,960 --> 00:18:39,480
你看我们可以自己过
547
00:18:40,599 --> 00:18:42,439
我们自己过的
548
00:18:43,599 --> 00:18:44,919
而没有采用高精地图的
549
00:18:44,960 --> 00:18:46,559
特斯拉则在根源上
550
00:18:46,559 --> 00:18:47,879
就不会有这种问题
551
00:18:54,440 --> 00:18:56,160
最后是人机共驾策略
552
00:18:56,400 --> 00:18:57,759
今年第一期 42Mark
553
00:18:57,759 --> 00:18:58,879
我们详细讲了
554
00:18:58,880 --> 00:19:00,400
人机共驾策略的重要性
555
00:19:00,720 --> 00:19:03,440
恰巧这次测试的 4 辆车都说过了
556
00:19:03,839 --> 00:19:04,879
大部分都是一样的
557
00:19:04,880 --> 00:19:05,560
强烈推荐
558
00:19:05,599 --> 00:19:07,559
没有看过的朋友回去看一看
559
00:19:07,880 --> 00:19:10,400
这是你和系统协作体验的关键
560
00:19:10,880 --> 00:19:12,280
在这里我也就不再重复了
561
00:19:12,480 --> 00:19:14,240
我们重点说一说基础辅助驾驶
562
00:19:14,279 --> 00:19:15,240
和导航辅助驾驶
563
00:19:15,279 --> 00:19:16,359
有区别的地方
564
00:19:16,359 --> 00:19:17,399
可视化方面
565
00:19:17,440 --> 00:19:19,200
特斯拉和蔚来在基础辅助驾驶
566
00:19:19,240 --> 00:19:21,039
和导航辅助驾驶状态下的体验
567
00:19:21,079 --> 00:19:22,399
基本上是一致的
568
00:19:22,680 --> 00:19:24,200
只有可视化上的车道线
569
00:19:24,200 --> 00:19:25,360
会从原来的车道
570
00:19:25,400 --> 00:19:26,560
变成一根引导线
571
00:19:26,839 --> 00:19:29,039
显示出车辆预计的行驶轨迹
572
00:19:29,440 --> 00:19:31,559
小鹏 P7 在这里逆风翻盘了
573
00:19:31,599 --> 00:19:33,879
直接把可视化拉高了一个维度
574
00:19:34,640 --> 00:19:36,120
在推出导航辅助驾驶同时
575
00:19:36,160 --> 00:19:37,240
小鹏联合高德
576
00:19:37,279 --> 00:19:38,799
推出了 SR 虚拟显示
577
00:19:38,839 --> 00:19:41,079
也就是辅助驾驶环境模拟显示
578
00:19:41,119 --> 00:19:42,759
不知道它这种车怎么识别
579
00:19:42,799 --> 00:19:44,039
就是在前面车一亮的
580
00:19:44,079 --> 00:19:45,599
SR建模的显示
581
00:19:45,640 --> 00:19:46,280
什么意思
582
00:19:46,519 --> 00:19:48,240
它是一个根据识别
583
00:19:48,279 --> 00:19:50,480
然后实时建模的一个东西
584
00:19:50,480 --> 00:19:52,880
前面车超过去它就会（建模显示）
585
00:19:52,880 --> 00:19:54,600
不仅可以基于传感器的感知
586
00:19:54,640 --> 00:19:56,280
显示周围车辆
587
00:19:56,279 --> 00:19:58,079
在高德高精地图的加持下
588
00:19:58,200 --> 00:19:59,880
你所在车道的环境
589
00:20:00,119 --> 00:20:02,239
对向车道的环境都一目了然
590
00:20:02,240 --> 00:20:03,720
效果非常震撼
591
00:20:04,319 --> 00:20:05,399
理想 ONE 同样采用了
592
00:20:05,400 --> 00:20:06,480
SR虚拟显示
593
00:20:06,920 --> 00:20:08,720
不过效果上略逊于小鹏
594
00:20:09,000 --> 00:20:10,519
一方面只有前向感知
595
00:20:10,519 --> 00:20:11,799
而且不会模拟显示
596
00:20:11,880 --> 00:20:13,760
毫米波雷达感知的结果
597
00:20:13,759 --> 00:20:15,799
周围车辆的显示没有小鹏多
598
00:20:15,839 --> 00:20:18,399
显示模型也不够小鹏那么细腻
599
00:20:18,799 --> 00:20:20,480
另一方面理想 ONE 的车机芯片
600
00:20:20,559 --> 00:20:22,480
需要同时支持中控屏
601
00:20:22,559 --> 00:20:24,240
和副驾两块屏幕
602
00:20:24,359 --> 00:20:26,319
所以副驾在看电视的时候
603
00:20:26,319 --> 00:20:27,839
SR显示可能会
604
00:20:27,839 --> 00:20:29,079
出现卡顿的情况
605
00:20:29,680 --> 00:20:31,960
总有一种为了上 SR 而上的感觉
606
00:20:32,519 --> 00:20:34,639
要么直接播电影得了
607
00:20:34,920 --> 00:20:37,160
反正感觉放这个被切进去不太好
608
00:20:37,440 --> 00:20:38,519
肯定不太好
609
00:20:38,559 --> 00:20:41,039
你可以直接放 42 号车库的视频
610
00:20:41,359 --> 00:20:42,240
看看自己
611
00:20:43,240 --> 00:20:45,279
可以看自己的视频
612
00:20:45,319 --> 00:20:48,039
然后瓜哥什么都行
613
00:20:48,039 --> 00:20:50,200
我们轮播
614
00:20:50,720 --> 00:20:51,600
导航辅助驾驶
615
00:20:51,599 --> 00:20:52,919
需要额外关注的一点
616
00:20:52,960 --> 00:20:54,200
就是变道确认
617
00:20:54,519 --> 00:20:56,279
上面我们提到了导航辅助驾驶
618
00:20:56,319 --> 00:20:58,119
具备自动变道的能力
619
00:20:58,160 --> 00:20:59,759
但是部分保守的车企
620
00:20:59,960 --> 00:21:02,160
在变道前会发起一个提醒
621
00:21:02,400 --> 00:21:04,160
只有经过驾驶员确认之后
622
00:21:04,200 --> 00:21:05,400
才会执行变道
623
00:21:05,759 --> 00:21:06,799
其中理想和特斯拉
624
00:21:06,960 --> 00:21:08,480
每一次变道都需要确认
625
00:21:08,480 --> 00:21:10,120
理想通过按键确认
626
00:21:10,160 --> 00:21:11,759
特斯拉通过轻扭方向盘
627
00:21:11,799 --> 00:21:13,839
或者是转向灯拨杆确认
628
00:21:14,079 --> 00:21:15,159
在操作的便利性上
629
00:21:15,319 --> 00:21:16,919
理想略好于特斯拉
630
00:21:17,240 --> 00:21:18,720
不过每一次变道确认
631
00:21:18,880 --> 00:21:21,160
不仅让变道的存在感极强
632
00:21:21,200 --> 00:21:22,319
确认的反应时间
633
00:21:22,440 --> 00:21:24,799
也有可能会错过最佳的变道时机
634
00:21:25,160 --> 00:21:26,960
而且每个变道场景下的缺点
635
00:21:27,240 --> 00:21:28,519
也都会被无限放大
636
00:21:28,799 --> 00:21:30,720
我同意这是出于安全的考虑
637
00:21:30,720 --> 00:21:33,960
但是背后确实是体验的山体滑坡
638
-->
这也是特斯拉变道
639
-->
积极让人反感的原因之一
640
00:21:42,960 --> 00:21:44,680
最后我们还统计了一个数据
641
00:21:45,039 --> 00:21:47,200
导航辅助驾驶的实际可用范围
642
00:21:47,519 --> 00:21:49,920
这是导航辅助驾驶功能的基础
643
00:21:50,359 --> 00:21:52,079
脱离了可用范围谈能力
644
00:21:52,079 --> 00:21:53,119
那都是耍流氓
645
00:21:53,160 --> 00:21:56,360
我觉得像这种江浙沪这边高速
646
00:21:56,400 --> 00:21:58,440
车流已经非常密集了
647
00:21:58,519 --> 00:21:59,279
大部分高速
648
00:21:59,279 --> 00:22:01,879
应该不会比这个再复杂了
649
00:22:02,279 --> 00:22:04,119
除非那种有事故堵车的
650
00:22:04,559 --> 00:22:06,399
所以我觉得能在这种高速上
651
00:22:06,799 --> 00:22:07,960
比较好用的话
652
00:22:08,000 --> 00:22:09,759
我觉得它的功能价值
653
00:22:09,799 --> 00:22:10,960
已经体现出来了
654
00:22:11,279 --> 00:22:12,200
最理想的状态下
655
00:22:12,200 --> 00:22:13,480
一辆车的导航辅助驾驶
656
00:22:13,559 --> 00:22:15,119
应该是 100% 可用的
657
00:22:15,519 --> 00:22:16,720
但是不可避免的是
658
00:22:16,759 --> 00:22:18,039
会出现驾驶员接管
659
00:22:18,039 --> 00:22:18,799
系统降级
660
00:22:18,839 --> 00:22:20,359
系统退出等等这些场景
661
00:22:20,920 --> 00:22:22,120
咱们不是走过吗
662
00:22:23,039 --> 00:22:24,240
回程可复杂了
663
00:22:29,480 --> 00:22:30,200
小鹏和理想
664
00:22:30,400 --> 00:22:32,040
因为高度依赖高精地图
665
00:22:32,079 --> 00:22:33,879
施工场景会直接退出
666
00:22:33,920 --> 00:22:36,240
同时在高精定位较弱的地方
667
00:22:36,440 --> 00:22:38,480
系统也出现了多次降级
668
00:22:38,480 --> 00:22:39,360
实际可用范围
669
00:22:39,680 --> 00:22:41,440
略逊于蔚来和特斯拉
670
00:22:42,720 --> 00:22:45,160
提前接管 NGP 即将退出
671
00:22:45,599 --> 00:22:47,959
退了 NGP 要退了
672
00:22:47,960 --> 00:22:48,680
NGP退了
673
00:22:49,000 --> 00:22:51,119
高速领航结束 100 m
674
00:22:56,920 --> 00:22:57,759
经过我们实测
675
00:22:57,759 --> 00:22:59,720
321公里的高速行程中
676
00:22:59,720 --> 00:23:01,440
特斯拉 蔚来实际可用范围最高
677
00:23:01,440 --> 00:23:03,840
达到了 99.5%
678
00:23:04,240 --> 00:23:06,720
也就是说这段路 99.5%
679
00:23:06,720 --> 00:23:08,319
都是车自己在开
680
00:23:08,759 --> 00:23:10,759
其次是小鹏 98.4%
681
00:23:10,759 --> 00:23:12,960
最后是理想 98.1%
682
00:23:15,960 --> 00:23:16,840
先让我们回顾一下
683
00:23:16,880 --> 00:23:19,440
这是这 4 辆车硬件能力的排名
684
00:23:19,720 --> 00:23:21,880
这是这 4 辆车实际体验的
685
00:23:21,880 --> 00:23:22,680
最终排名
686
00:23:24,599 --> 00:23:25,919
硬件能力上限最高的
687
00:23:25,920 --> 00:23:27,360
特斯拉排名垫底
688
00:23:27,400 --> 00:23:29,200
从实测的变道场景
689
00:23:29,240 --> 00:23:30,200
特殊场景来看
690
00:23:30,680 --> 00:23:33,039
特斯拉的绝对能力是最强的
691
00:23:33,279 --> 00:23:35,599
排名垫底的主要原因是策略
692
00:23:35,960 --> 00:23:37,480
无论是变道还是匝道的体验
693
00:23:37,640 --> 00:23:38,840
都是惊心动魄的
694
00:23:38,880 --> 00:23:41,000
可以说是不懂中国道路环境
695
00:23:41,240 --> 00:23:43,160
也可以说是压根没用心标定
696
00:23:43,200 --> 00:23:45,240
这也导致了很多人买了
697
00:23:45,279 --> 00:23:46,559
导航辅助驾驶的软件
698
00:23:46,559 --> 00:23:48,839
但是更愿意用基础辅助驾驶
699
00:23:49,279 --> 00:23:50,039
如果你也是这样的
700
00:23:50,079 --> 00:23:51,119
特斯拉车主
701
00:23:51,160 --> 00:23:52,440
可以弹幕里扣个 1
702
00:23:52,440 --> 00:23:53,440
变道
703
00:23:53,440 --> 00:23:54,680
倒是一气呵成
704
00:23:54,680 --> 00:23:56,519
但是就是老是做一些
705
00:23:56,519 --> 00:23:58,599
无效变道的事情出来就很奇怪
706
00:23:58,599 --> 00:24:00,359
最晚发布的理想 ONE 拿到了
707
00:24:00,359 --> 00:24:01,599
刚刚及格的成绩
708
00:24:01,640 --> 00:24:03,720
这并不是一个非常出色的表现
709
00:24:03,759 --> 00:24:04,759
但是在这背后
710
00:24:04,799 --> 00:24:06,240
理想的自动驾驶研发
711
00:24:06,440 --> 00:24:08,360
从老款的理想 ONE 的供应商为主
712
00:24:08,440 --> 00:24:09,960
转换到了全栈自研
713
00:24:10,000 --> 00:24:12,440
整体节奏上相比大家慢了一步
714
00:24:12,440 --> 00:24:14,039
只能说新款理想 ONE 的表现
715
00:24:14,119 --> 00:24:15,319
是一个开始
716
00:24:15,319 --> 00:24:16,240
工作的稳定性
717
00:24:16,279 --> 00:24:17,799
以及感知控制的能力
718
00:24:17,960 --> 00:24:19,319
都有提升的空间
719
00:24:19,640 --> 00:24:21,120
最终能做到怎么样的水平
720
00:24:21,160 --> 00:24:22,840
我们还需要继续观察
721
00:24:23,279 --> 00:24:25,480
变啊大哥
722
00:24:25,519 --> 00:24:26,960
刹车变
723
00:24:34,559 --> 00:24:41,759
变呀大哥
724
00:24:47,039 --> 00:24:48,440
终于等到你
725
00:24:49,480 --> 00:24:52,039
还好我没放弃
726
00:24:52,519 --> 00:24:53,839
硬件能力垫底的蔚来
727
00:24:53,920 --> 00:24:54,880
在实际体验上
728
00:24:55,079 --> 00:24:57,079
可以说是名副其实的第二
729
00:24:57,079 --> 00:24:58,919
相比一年前的变道不积极
730
00:24:58,960 --> 00:25:00,319
匝道肌无力的表现
731
00:25:00,480 --> 00:25:03,000
可以说是有了翻天覆地的变化
732
00:25:03,000 --> 00:25:05,200
2020年末的时候
733
00:25:05,200 --> 00:25:07,720
不对 2021 年初的时候
734
00:25:07,720 --> 00:25:11,039
小鹏刚推这个功能当时
735
00:25:11,039 --> 00:25:13,079
推了之后第一次用
736
00:25:13,079 --> 00:25:16,519
我感觉这不就是自动驾驶吗
737
00:25:16,519 --> 00:25:19,879
现在在我心里蔚来就是这样
738
00:25:19,880 --> 00:25:24,280
蔚来真的是把毫米波用到了极限
739
00:25:24,279 --> 00:25:26,000
用毫米波实现了视觉的效果
740
00:25:26,000 --> 00:25:28,759
下午这一路坐得多顺利啊
741
00:25:28,759 --> 00:25:30,440
这套 17 年定义的硬件架构
742
00:25:30,480 --> 00:25:32,559
能被蔚来压榨到现在的表现
743
00:25:32,599 --> 00:25:34,399
着实是非常惊艳了
744
00:25:34,400 --> 00:25:36,560
而且蔚来虽然也用了高精地图
745
00:25:36,599 --> 00:25:38,359
但是在执行上仍然采用了
746
00:25:38,400 --> 00:25:39,560
技术难度更高的
747
00:25:39,599 --> 00:25:40,919
视觉为主路线
748
00:25:41,400 --> 00:25:43,600
这样的进步幅度和硬件压榨能力
749
00:25:43,640 --> 00:25:44,880
也让我们无限期待
750
00:25:44,920 --> 00:25:46,800
全新硬件架构的 ET7
751
00:25:46,839 --> 00:25:48,199
能做到怎么样的水平
752
00:25:48,839 --> 00:25:50,119
小鹏 P7 极低的
753
00:25:50,119 --> 00:25:51,279
百公里接管次数
754
00:25:51,319 --> 00:25:52,679
和稳定的匝道表现
755
00:25:52,759 --> 00:25:54,160
给人极强的自动感
756
00:25:54,440 --> 00:25:55,559
但是需要批评的是
757
00:25:55,599 --> 00:25:57,399
在车流相对较大的环境中
758
00:25:57,440 --> 00:25:58,559
小鹏的变道策略
759
00:25:58,599 --> 00:26:00,519
仍然有比较大的优化空间
760
00:26:00,759 --> 00:26:02,759
而且非常遗憾的是过去一年
761
00:26:02,799 --> 00:26:04,279
小鹏的导航辅助驾驶能力
762
00:26:04,319 --> 00:26:06,319
似乎并没有太大的提升
763
00:26:06,319 --> 00:26:10,200
这一路上 NGP 给我带来太多惊喜
764
00:26:10,200 --> 00:26:11,039
虽然有一点惊吓
765
00:26:11,039 --> 00:26:12,519
但大多数都是惊喜
766
00:26:12,519 --> 00:26:16,759
而且确实能够减低很多的驾驶疲劳
767
00:26:16,759 --> 00:26:18,680
毕竟它这个路线都给你规划好了
768
00:26:18,680 --> 00:26:20,360
你基本上只要看着路就可以了
769
00:26:21,000 --> 00:26:22,119
不可否认的是小鹏
770
00:26:22,119 --> 00:26:23,959
仍然是综合体验的王者
771
00:26:23,960 --> 00:26:25,400
总体来说小鹏这个
772
00:26:25,400 --> 00:26:28,120
无效变道的概率不高
773
00:26:29,000 --> 00:26:30,400
其实大部分的变道
774
00:26:30,440 --> 00:26:32,120
还是有效变道
775
00:26:32,160 --> 00:26:34,160
就是直接变到更快的车道上去
776
00:26:34,200 --> 00:26:36,200
它这次整体的体验下来
777
00:26:36,200 --> 00:26:37,720
我觉得小鹏的 NGP
778
00:26:37,720 --> 00:26:39,920
给我的信心非常的足 很爽
779
00:26:39,920 --> 00:26:41,600
但是就现在的表现来看
780
00:26:41,720 --> 00:26:44,240
远没有达到这套硬件的能力上限
781
00:26:44,240 --> 00:26:46,559
所以也希望小鹏的辅助驾驶团队
782
00:26:46,720 --> 00:26:47,839
能够再接再厉
783
00:26:47,880 --> 00:26:50,160
给 P7 车主们多一些惊喜
784
00:26:50,480 --> 00:26:51,559
小鹏厉害的地方
785
00:26:51,559 --> 00:26:53,000
就是它变道真的很稳
786
00:26:53,000 --> 00:26:54,680
而且它没有什么无效变道
787
00:26:54,680 --> 00:26:55,360
有没有发现
788
00:26:55,400 --> 00:26:56,880
它基本上所有变道
789
00:26:56,920 --> 00:26:58,320
都是能给你一个
790
00:26:58,359 --> 00:26:59,439
虽然它有的时候比较慢
791
00:26:59,440 --> 00:27:01,360
但是它至少变道的成功率是很高的
792
00:27:01,359 --> 00:27:02,919
从 ACC 到基础辅助驾驶
793
00:27:02,960 --> 00:27:04,440
再到导航辅助驾驶
794
00:27:04,480 --> 00:27:06,799
车的自动化程度越来越高
795
00:27:06,839 --> 00:27:08,399
能力范围越来越广
796
00:27:08,519 --> 00:27:10,359
但是争议也越来越大
797
00:27:10,680 --> 00:27:12,920
而这些争议主要集中在部分车型
798
00:27:12,920 --> 00:27:15,840
实际表现太差和功能太过期货上
799
00:27:16,359 --> 00:27:18,399
而就我们这次实测的表现来看
800
00:27:18,440 --> 00:27:19,880
在相对通畅的高速
801
00:27:19,880 --> 00:27:20,680
最强的小鹏呢
802
00:27:20,720 --> 00:27:22,880
已经可以做到百公里一次接管了
803
00:27:23,200 --> 00:27:24,080
而且我们还看到
804
00:27:24,119 --> 00:27:26,159
今年即将量产的车型都换上了
805
00:27:26,200 --> 00:27:27,759
更加强大的传感器
806
00:27:28,160 --> 00:27:29,120
越来越多的厂家
807
00:27:29,160 --> 00:27:31,400
也开始投入大量的人力资金
808
00:27:31,440 --> 00:27:32,640
研发自动驾驶
809
00:27:32,960 --> 00:27:35,120
自动驾驶不会从天上直接掉下来
810
00:27:35,119 --> 00:27:36,599
而导航辅助驾驶
811
00:27:36,880 --> 00:27:37,840
一定是我们实现
812
00:27:37,880 --> 00:27:40,040
全自动驾驶的必经之路
813
00:27:40,680 --> 00:27:41,840
OK以上就是本期
814
00:27:41,880 --> 00:27:43,040
视频的全部内容了
815
00:27:43,079 --> 00:27:44,960
这期视频可以说是又干又硬
816
00:27:45,200 --> 00:27:46,960
超级感谢你能够看到这里
817
00:27:47,200 --> 00:27:48,920
如果你觉得这期视频还不错的话
818
00:27:49,039 --> 00:27:50,839
一定要给我们来一个一键三连
819
00:27:51,119 --> 00:27:52,039
这是是 42 号车库
820
00:27:52,079 --> 00:27:52,839
我是63
821
00:27:53,319 --> 00:27:54,000
对了
822
00:27:54,000 --> 00:27:56,079
我们对导航辅助驾驶能力的探索
823
00:27:56,079 --> 00:27:57,000
不只在高速
824
00:27:57,039 --> 00:27:59,200
所以我们还在高架进行了测试
825
00:27:59,440 --> 00:28:01,000
这一期视频正在策划中
826
00:28:01,000 --> 00:28:03,039
已经有了突破性的进展了
827
00:28:03,039 --> 00:28:04,759
欢迎持续关注
50789
