WEBVTT
Kind: captions
Language: en

00:00:00.570 --> 00:00:05.090
A robot is just outside a warehouse and needs
to get to the second door.

00:00:05.090 --> 00:00:06.800
It can move on a line.

00:00:06.800 --> 00:00:09.360
However, it doesn't know where it is.

00:00:09.360 --> 00:00:11.599
How does it localize itself?

00:00:11.599 --> 00:00:17.039
It has wheel odometry, a camera to detect
a door, and a map of the warehouse.

00:00:17.039 --> 00:00:20.820
We can model the belief on where the robot
is as a probability distribution over the

00:00:20.820 --> 00:00:23.140
possible locations.

00:00:23.140 --> 00:00:28.680
Just deployed, the probability is uniform
as the robot's initial location is unknown.

00:00:28.680 --> 00:00:32.509
As the robot gets an image from the camera,
it detects a door.

00:00:32.509 --> 00:00:37.180
Given that the map shows there are three doors,
the probability increases at the locations

00:00:37.180 --> 00:00:39.210
in front of each door.

00:00:39.210 --> 00:00:44.160
The robot moves right, and the belief is also
updated accordingly, with the information

00:00:44.160 --> 00:00:46.270
from the wheel odometry.

00:00:46.270 --> 00:00:50.530
The robot then senses a second door and the
probability corresponding to location at the

00:00:50.530 --> 00:00:53.570
second door becomes the highest, localizing
the robot.