1
00:00:00,860 --> 00:00:01,250
Okay.

2
00:00:01,520 --> 00:00:05,540
Before diving into the actual motion planning, let's address the elephant in the room.

3
00:00:05,960 --> 00:00:09,470
We don't have the robot orientation, so how are we going to get it?

4
00:00:09,920 --> 00:00:11,990
Let's ask the board itself that.

5
00:00:11,990 --> 00:00:13,220
What is it looking at?

6
00:00:14,360 --> 00:00:18,260
So addressing the query, we gather what we already know about the robot.

7
00:00:18,590 --> 00:00:21,260
We have its location at each iteration.

8
00:00:21,620 --> 00:00:24,020
But how do we get its orientation?

9
00:00:24,290 --> 00:00:27,230
We found the robot location using background subtraction.

10
00:00:27,590 --> 00:00:31,040
We identified the foreign object and detected its location.

11
00:00:31,340 --> 00:00:37,130
But for robot orientation, we require some extra information about the robot structure so that we could

12
00:00:37,130 --> 00:00:40,070
identify what is facing at any moment.

13
00:00:40,820 --> 00:00:46,010
Asking the original question to the robot, we can get some related information out of it that what

14
00:00:46,010 --> 00:00:48,440
is its orientation in the simulation?

15
00:00:49,550 --> 00:00:56,000
But we require its orientation in the image because of consistency of frame of reference as the location

16
00:00:56,000 --> 00:00:58,070
extracted earlier was in the image.

17
00:00:58,730 --> 00:01:03,760
So this means we need to find some relationship between the orientation and simulation with the car

18
00:01:03,830 --> 00:01:04,820
orientation image.

19
00:01:05,360 --> 00:01:11,930
If we do that, we could use this relationship to compute the car orientation image from the car orientation

20
00:01:11,930 --> 00:01:12,800
in simulation.

21
00:01:13,700 --> 00:01:16,580
Let's look at the algorithm that does exactly that.

22
00:01:17,780 --> 00:01:21,080
You start off by extracting the orientation and simulation.

23
00:01:21,470 --> 00:01:27,180
Then we compute the orientation image by moving the car forward for a few frames and then computing

24
00:01:27,180 --> 00:01:28,850
the angle using trigonometry.

25
00:01:29,480 --> 00:01:34,220
Then we find offset between the patient's simulation and the orientation image.

26
00:01:35,000 --> 00:01:41,420
Finally, we use this compute offset to transform the orientation simulation to the orientation in image.

27
00:01:42,290 --> 00:01:47,660
Once we do that, we now have the complete robot pose both its location and orientation.

28
00:01:48,050 --> 00:01:51,800
This means that we could move forward to define the go to function.