All language subtitles for 01_map-registration.en

af Afrikaans
sq Albanian
am Amharic
ar Arabic Download
hy Armenian
az Azerbaijani
eu Basque
be Belarusian
bn Bengali
bs Bosnian
bg Bulgarian
ca Catalan
ceb Cebuano
ny Chichewa
zh-CN Chinese (Simplified)
zh-TW Chinese (Traditional)
co Corsican
hr Croatian
cs Czech
da Danish
nl Dutch
en English
eo Esperanto
et Estonian
tl Filipino
fi Finnish
fr French
fy Frisian
gl Galician
ka Georgian
de German
el Greek
gu Gujarati
ht Haitian Creole
ha Hausa
haw Hawaiian
iw Hebrew
hi Hindi
hmn Hmong
hu Hungarian
is Icelandic
ig Igbo
id Indonesian
ga Irish
it Italian
ja Japanese
jw Javanese
kn Kannada
kk Kazakh
km Khmer
ko Korean
ku Kurdish (Kurmanji)
ky Kyrgyz
lo Lao
la Latin
lv Latvian
lt Lithuanian
lb Luxembourgish
mk Macedonian
mg Malagasy
ms Malay
ml Malayalam
mt Maltese
mi Maori
mr Marathi
mn Mongolian
my Myanmar (Burmese)
ne Nepali
no Norwegian
ps Pashto
fa Persian
pl Polish
pt Portuguese
pa Punjabi
ro Romanian
ru Russian
sm Samoan
gd Scots Gaelic
sr Serbian
st Sesotho
sn Shona
sd Sindhi
si Sinhala
sk Slovak
sl Slovenian
so Somali
es Spanish
su Sundanese
sw Swahili
sv Swedish
tg Tajik
ta Tamil
te Telugu
th Thai
tr Turkish
uk Ukrainian
ur Urdu
uz Uzbek
vi Vietnamese
cy Welsh
xh Xhosa
yi Yiddish
yo Yoruba
zu Zulu
or Odia (Oriya)
rw Kinyarwanda
tk Turkmen
tt Tatar
ug Uyghur
Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated: 1 00:00:05,010 --> 00:00:10,224 In this lecture, we will consider correlation based matching strategies for 2 00:00:10,224 --> 00:00:13,470 location a robot on a map given laser range data. 3 00:00:14,500 --> 00:00:19,550 This map registration process provides a very precise complement 4 00:00:19,550 --> 00:00:21,570 to odometry based localization. 5 00:00:22,920 --> 00:00:26,140 First we should introduced the LIDAR depth sensor. 6 00:00:27,200 --> 00:00:30,710 LIDAR stands for light detection and ranging and 7 00:00:30,710 --> 00:00:32,690 it provides distance measurements. 8 00:00:32,690 --> 00:00:36,600 Often engineered in a laser scanner to provide two dimensional data. 9 00:00:38,540 --> 00:00:41,310 The laser scanner we will model in this lecture 10 00:00:41,310 --> 00:00:44,560 takes depth measurements in polar coordinates, 11 00:00:44,560 --> 00:00:49,500 where a continuous distance reading r is made at discrete angles theta. 12 00:00:50,530 --> 00:00:55,660 Here, theta encompasses 270 degrees, not a full circle. 13 00:00:57,210 --> 00:01:00,420 The laser scanner can only see 10 to 30 meters away. 14 00:01:01,610 --> 00:01:05,000 In this range restriction, means that distance measurements 15 00:01:05,000 --> 00:01:09,800 showing here is black dots, can only be found within the area in green. 16 00:01:10,910 --> 00:01:16,320 Thus, due to the rays generating from a single point and the limited range, 17 00:01:16,320 --> 00:01:20,210 the robot can only see the dotted lines and not the lines in brown. 18 00:01:22,150 --> 00:01:27,160 Just as in the previous lecture, a two dimensional occupancy grid map will 19 00:01:27,160 --> 00:01:32,090 be used in localization, where a light colored cell represents high 20 00:01:32,090 --> 00:01:37,440 probability of an obstacle and a dark colored cell present a low probability. 21 00:01:38,620 --> 00:01:41,960 The cells here are meant to replicate the laser skin 22 00:01:41,960 --> 00:01:46,830 shown in the previous slide when the robot is approaching a corner in a hallway. 23 00:01:48,640 --> 00:01:54,350 Because the robot lives in a finite grid world the grid must sometimes be expanded 24 00:01:54,350 --> 00:01:56,110 as the robot can escape the boundaries. 25 00:01:57,140 --> 00:02:01,730 In this case, the map representation should increase in size as the robot 26 00:02:01,730 --> 00:02:06,720 turns and travels on the corridor shown in the top left of this map or 27 00:02:06,720 --> 00:02:08,520 else information will be lost. 28 00:02:10,050 --> 00:02:14,310 In addition to mapping the laser data discussed in week 3, we can 29 00:02:14,310 --> 00:02:18,920 access map data and try to find the robot pose in the map given the laser data. 30 00:02:20,480 --> 00:02:25,990 The complimentary stages of mapping and localization when performed together 31 00:02:25,990 --> 00:02:29,170 are known as SLAM, simultaneous localization and 32 00:02:29,170 --> 00:02:32,399 mapping, which is a major research topic in robotics. 33 00:02:34,190 --> 00:02:37,110 In the localization problem we have two sets of information. 34 00:02:38,160 --> 00:02:42,280 First, the occupancy grid map provides a grounds truth knowledge 35 00:02:42,280 --> 00:02:44,590 of what the robot should expect to observe in the world. 36 00:02:45,730 --> 00:02:48,710 Second, the set of lighter scan measurements 37 00:02:48,710 --> 00:02:52,240 provides information on what the robot is observing at the current time. 38 00:02:53,780 --> 00:02:56,580 The lighter scan measurements must be discretized 39 00:02:56,580 --> 00:03:00,860 according to the map representation, as discussed in week three, 40 00:03:00,860 --> 00:03:03,880 in order to be compared to the information from the occupancy map. 41 00:03:05,750 --> 00:03:10,680 With these two pieces of information the goal is to find the best robot pose 42 00:03:10,680 --> 00:03:13,289 on the map that explains the measured observations. 43 00:03:14,950 --> 00:03:19,100 Searching over all possible poses of the robot can be difficult. 44 00:03:19,100 --> 00:03:22,350 But based on the odometry information discussed in the last lecture, 45 00:03:22,350 --> 00:03:24,800 we have some tricks to make the search easier. 46 00:03:26,680 --> 00:03:29,600 We can constrain the search to a limited number of poses 47 00:03:29,600 --> 00:03:30,990 based on odometry information. 48 00:03:32,050 --> 00:03:37,310 Because we track the robot over time, we have the last known position of the robot 49 00:03:37,310 --> 00:03:40,780 and odometry information on how far the robot most likely moved. 50 00:03:42,120 --> 00:03:47,090 Thus, the most likely pose for the robot is now given a new set of laser data, 51 00:03:47,090 --> 00:03:50,310 is probably close to where the odometry predicts the robot to be. 52 00:03:51,360 --> 00:03:55,750 This prediction means that we can refine our search to poses near the prediction 53 00:03:55,750 --> 00:03:58,420 and be more confident in the validity of our search results. 54 00:04:00,150 --> 00:04:05,070 We measure each pose p in the search based on a map registration metric. 55 00:04:06,120 --> 00:04:10,910 One metric is to consider the sum of the map values m, at coordinates x and y, 56 00:04:10,910 --> 00:04:13,320 where the laser returns r, hit. 57 00:04:14,970 --> 00:04:19,270 This correlation metric can be modified to suit the application at hand. 58 00:04:19,270 --> 00:04:23,200 In our case, the value of our map cell will be a log odds ratio, so 59 00:04:23,200 --> 00:04:27,090 laser returns that are seen at a map location with high probability of 60 00:04:27,090 --> 00:04:31,840 occupancy will strongly increase the registration in the metric score. 61 00:04:33,190 --> 00:04:38,590 Laser returns with map locations known as free cells will decrease the metric score. 62 00:04:39,610 --> 00:04:44,140 Additionally, the correlation can be scaled where returns from far 63 00:04:44,140 --> 00:04:48,710 distances affect the metric calculation less than nearby the laser returns. 64 00:04:50,350 --> 00:04:52,240 We register the robot on the map, 65 00:04:52,240 --> 00:04:54,740 at the pose that maximizes the registration metric. 66 00:04:55,760 --> 00:04:57,670 Thus, when the odometry is calculated, 67 00:04:57,670 --> 00:05:00,870 it uses this pose to predict a new position of the robot, in time. 68 00:05:02,830 --> 00:05:06,930 In addition to considering merely the laser returns, we can consider points for 69 00:05:06,930 --> 00:05:09,460 the laser returns penetrated. 70 00:05:09,460 --> 00:05:12,380 This calculation can further corroborate our map registration. 71 00:05:13,630 --> 00:05:15,910 It requires considerably more computation however. 72 00:05:17,240 --> 00:05:21,180 To capture pose uncertainty using a simple Gaussian on position and 73 00:05:21,180 --> 00:05:24,430 angle may not provide a feasible approach. 74 00:05:24,430 --> 00:05:28,700 In the next lecture, we will present a pose filter that can capture bi-modal 75 00:05:28,700 --> 00:05:32,917 uncertainty and non-linear models and a computationally tractable way.7184

Can't find what you're looking for?
Get subtitles in any language from opensubtitles.com, and translate them here.