WEBVTT

00:00.000 --> 00:10.000
Hello, everyone.

00:10.000 --> 00:13.000
Welcome to the Lampking Presentation.

00:13.000 --> 00:14.000
My name is Llamar Sal.

00:14.000 --> 00:17.000
I'm a robotics software developer at the command.

00:17.000 --> 00:19.000
And last year we bring here Beluga.

00:19.000 --> 00:22.000
Beluga is an MSEL.

00:22.000 --> 00:25.000
Montecal localization framework for robotics.

00:25.000 --> 00:29.000
So this year I'm pleased to bring here it's sibling.

00:29.000 --> 00:33.000
It's about how to make much marketing to the work for the organization.

00:33.000 --> 00:35.000
So what is Lampking?

00:35.000 --> 00:38.000
Lampking is localization and mapping with machine for robotics.

00:38.000 --> 00:41.000
And it is based on these four pylars.

00:41.000 --> 00:43.000
So let's go straight on the pylars.

00:43.000 --> 00:46.000
The architecture depends on the orchestration.

00:46.000 --> 00:48.000
So first of all, reproduce all the environments.

00:48.000 --> 00:50.000
Because it's based on your contents.

00:50.000 --> 00:51.000
It's based on earthly.

00:51.000 --> 00:54.000
At least a mix between Docker and make.

00:54.000 --> 00:58.000
Also the another pylars is standard performance metrics.

00:58.000 --> 01:01.000
So it contains EVO type memory.

01:01.000 --> 01:05.000
Both are like tools for metrics.

01:05.000 --> 01:09.000
So EVO gives you the absolute post error and the relative post error.

01:09.000 --> 01:13.000
And the memory is really good for performance benchmarking.

01:13.000 --> 01:19.000
And also we have to draw two that gives us like there was like play and recording.

01:19.000 --> 01:20.000
And so on.

01:20.000 --> 01:22.000
So it's it's really used to integrate it.

01:22.000 --> 01:24.000
And other pylars.

01:24.000 --> 01:26.000
The collaborative benchmark of the finishes.

01:26.000 --> 01:29.000
So it contains also it was on the framework.

01:29.000 --> 01:35.000
That it helps us to like to have a link between or a hook between the keyword.

01:35.000 --> 01:37.000
And the phrase itself.

01:37.000 --> 01:42.000
So it's it's really easy to write the sketches or that is bench.

01:42.000 --> 01:45.000
So and finally for that fourth pylar.

01:45.000 --> 01:47.000
Automatic representation.

01:47.000 --> 01:49.000
We are using LATX.

01:49.000 --> 01:50.000
So let's go.

01:50.000 --> 01:52.000
So this is architecture.

01:53.000 --> 01:55.000
So one of the shepherd part.

01:55.000 --> 01:59.000
You can see mostly on the interaction between the robot framework.

01:59.000 --> 02:04.000
With like gives us this easy way to to write.

02:04.000 --> 02:06.000
That test benchmarks.

02:06.000 --> 02:08.000
And the Lampking clerk is about more.

02:08.000 --> 02:10.000
How we want to reproduce.

02:10.000 --> 02:11.000
How do I want to plot.

02:11.000 --> 02:15.000
How do I want to get the data and report it in LATX report.

02:15.000 --> 02:19.000
So let's go to try in a use case.

02:19.000 --> 02:21.000
So as I said before.

02:21.000 --> 02:24.000
Let's start what we bring here Benuja.

02:24.000 --> 02:27.000
So it's a Monte Carlo calculation framework.

02:27.000 --> 02:29.000
So these are.

02:29.000 --> 02:34.000
These are the configurations parameters that this bench contain.

02:34.000 --> 02:37.000
But you can use whatever you want for your application.

02:37.000 --> 02:40.000
So for example, number of particles and some model kinematic model.

02:40.000 --> 02:42.000
It depends on your application.

02:42.000 --> 02:48.000
Also you will have different topics that can use as an input for your application.

02:48.000 --> 02:52.000
And lower post that is going to be what we contain.

02:52.000 --> 02:56.000
So what happens in localization when you.

02:56.000 --> 02:59.000
When you try to to get localization from a robot.

02:59.000 --> 03:04.000
Sometimes you get like because it's a stochastically algorithm.

03:04.000 --> 03:06.000
You can get that you are here.

03:06.000 --> 03:07.000
Sometimes.

03:07.000 --> 03:09.000
So you cannot run it once.

03:09.000 --> 03:11.000
Just to know that it's work as expected.

03:11.000 --> 03:13.000
You have to run multiple times.

03:13.000 --> 03:16.000
That's why you need to do like.

03:16.000 --> 03:20.000
This bench and to get different kind of test just to see how it works.

03:20.000 --> 03:22.000
So how it works.

03:22.000 --> 03:24.000
You have just.

03:24.000 --> 03:25.000
Like fine.

03:25.000 --> 03:27.000
That is this one.

03:27.000 --> 03:28.000
And here you have.

03:28.000 --> 03:29.000
The setup.

03:29.000 --> 03:33.000
So we're going to run this the beginning of our test bench.

03:33.000 --> 03:37.000
Then we're going to like run this at the end of our bench.

03:37.000 --> 03:38.000
Or this bench.

03:38.000 --> 03:41.000
And finally we're going to like.

03:41.000 --> 03:43.000
This is going to be the test template itself.

03:43.000 --> 03:45.000
So you can see that we can.

03:45.000 --> 03:47.000
Like right really easily different.

03:47.000 --> 03:49.000
Like by the voice.

03:49.000 --> 03:52.000
And let's go try to the test template.

03:52.000 --> 03:53.000
Sorry.

03:53.000 --> 03:54.000
No.

03:54.000 --> 03:55.000
Let's go to the by the voice.

03:55.000 --> 03:57.000
So as you can see here we have.

03:57.000 --> 04:00.000
We can do the test bench for two different laser models.

04:00.000 --> 04:01.000
For the beam.

04:01.000 --> 04:03.000
I'm from the likelihood field.

04:03.000 --> 04:06.000
And we can also run this.

04:06.000 --> 04:10.000
Using this specific laser models to different datasets.

04:10.000 --> 04:14.000
So let's go try to the test bench.

04:14.000 --> 04:16.000
So it is easy.

04:16.000 --> 04:18.000
You are writing the different.

04:18.000 --> 04:19.000
It's like syntax.

04:19.000 --> 04:21.000
You are writing whatever you want.

04:21.000 --> 04:22.000
You are.

04:22.000 --> 04:23.000
You are telling. Okay.

04:23.000 --> 04:24.000
Let's set this viable.

04:24.000 --> 04:25.000
Let's use this dataset.

04:25.000 --> 04:27.000
But so everything is dynamically.

04:27.000 --> 04:29.000
So you can use different.

04:29.000 --> 04:30.000
Data sets.

04:30.000 --> 04:32.000
Different configuration parameters.

04:32.000 --> 04:35.000
And you write the test case.

04:35.000 --> 04:36.000
And.

04:36.000 --> 04:39.000
Finally when you have all the tests.

04:39.000 --> 04:41.000
You're running this multiple times.

04:41.000 --> 04:43.000
For example, we want to run 10 times.

04:43.000 --> 04:46.000
And finally in the.

04:46.000 --> 04:50.000
Um, tear down what we are doing is getting the report itself.

04:50.000 --> 04:52.000
So this is the report.

04:52.000 --> 04:54.000
Feel free to catch.

04:54.000 --> 04:57.000
And this is one of examples on every very value.

04:57.000 --> 04:58.000
A release.

04:58.000 --> 04:59.000
We release one of the report.

04:59.000 --> 05:00.000
So it's nice.

05:00.000 --> 05:02.000
And we are looking.

05:02.000 --> 05:04.000
We are going to start refactoring this.

05:04.000 --> 05:05.000
Um, report.

05:05.000 --> 05:07.000
Because there are some of the packages.

05:07.000 --> 05:08.000
That's been a long time.

05:08.000 --> 05:10.000
So we're really open to.

05:10.000 --> 05:11.000
To talk with you.

05:11.000 --> 05:13.000
And feel free to download.

05:13.000 --> 05:14.000
For.

05:14.000 --> 05:15.000
For.

05:15.000 --> 05:16.000
Um, give stars.

05:16.000 --> 05:18.000
And we are open to.

05:18.000 --> 05:19.000
Like come.

05:19.000 --> 05:20.000
Like to talk about.

05:20.000 --> 05:21.000
And.

05:21.000 --> 05:22.000
And let's do together.

05:22.000 --> 05:25.000
The framework for local decision that we deserve.

05:25.000 --> 05:26.000
Thank you.

05:26.000 --> 05:36.000
Thank you.