Virtual Gothenburg | Build: London ’19 for Architecture | Unreal Engine

Virtual Gothenburg | Build: London ’19 for Architecture | Unreal Engine


>>We are trying
a new concept here, starting with a film. Thanks for inviting us
here to the Epic Games, to this Epic Build. It is very inspiring to be here
in front of this crowd. My name is Eric Jeansson,
and I am a Geo Data Strategist, the city of Gothenburg
in Sweden. And this is Harald Karlsson,
engineer. And we are going
to talk about, tell the story about our digital
twin, Virtual Gothenburg. That, you have already seen. How do we face
the challenge of building a smart and sustainable city, with new ways of living,
new ways of meeting, new ways of transportation,
and new energy solutions? How do we face the challenges
of climate change, segregation,
the new transport needs, and the amount raising, raising data amount
in the society? One way is to build
a digital twin. We have seen
an example here earlier. We called it Virtual Gothenburg,
and it will be finished in 2021, when we have our 400 years
anniversary of Gothenburg. You are welcome
to visit us then. A digital twin is not just
a visual copy of reality, it is also the information
which is connected to it. This is very important. And in Gothenburg, we see,
as a smart city, we share data with each other, more processes
in the city share data. And in central, this circle,
we have Virtual Gothenburg, which will act as a platform
for that kind of data. We can visualize, analyze,
and demonstrate. Okay, you can divide, you say,
the data in the digital twin, and then what you can use
this digital twin for. You can first talk
about the present data, which could be used
to understand the city, how it works. And then we have real-time data
you can use to control the city, like the traffic
and flow of people in town, pollutions and so on. And for the third,
you have simulated data, or scenario data, which you
can use to predict the future. We have large investments, to say about 10 million Euro
in West Sweden concerning this kind of digital
twin technology. And one project that is led by
Chalmers Technical University, it is about building
a platform based on Unreal Engine
for digital twins. And Virtual Gothenburg,
our digital twin, will serve as base
for that kind of research. But now to
the tricky question, Harald, how do we make this happen?>>Yes, that is
the tricky question indeed. So, my name is Harald Karlsson. I work on this Virtual
Gothenburg project as bringing it into Unreal,
and how that pipeline will look. So, the basic concept
of our pipeline is we start with
all this data, right? There is data scattered
all over the city. So, some people
have the roads, but someone else
has the traffic lights, and some data does not
even exist that we need. For example,
we did not have database covering rooftop materials
for all the buildings, and that is kind of essential
to building visually pleasing model
of the city. So, we have to generate
that ourselves by looking at photographic
images of the city, and building the AI that looked
at all these rooftops and calculating what kind
of materials they can use, and stuff like that. So, you get all these kinds
of different data held by different people, and the next step
from collecting that data is to process it into a database
that we can control, and actually look
at all the data and have it in one place, we can actually use it
for something useful, right? That data is then pushed
into ESRI’s CityEngine that was mentioned before
to generate procedural models. So, if you have the attributes
of the number of floors, how the windows look,
when the building was built and all the kinds
of interesting data, you can put that into a 2D space
and generate the building. Very simple,
and very interesting. But there is a lot
of buildings in the city, so this requires a way of
looking at them at the same time, right? And that is where Unreal
comes into the picture. So, Unreal system will bring
all these assets with LOD levels,
collision boxes, all the good stuff, and put it together
so you can actually look at the virtual version
of the city in real-time. Then people are able
to use this digital twin to train autonomous vehicles, or may flood it
with a bunch of water, or any kind of innovation, science thing
you can come up with. They can use a virtual
copy of the city. So, to do all this,
we use World Composition. It is an Unreal feature
that basically splits up the map into a grid to get these tiles. And this is one of those tiles.
It is one kilometer across, and I want to focus
on this colorful, happy texture over here. We called it
semantic segmentation tile or something cool
like that. If you have a cool name for it,
you can poke me later. This texture can contain
thousands of data points, because we can segment
the colors, so dark green can be
where to spawn mushrooms, and light green can be
where to spawn blueberries. But the real bottleneck is
to actually have a database that contains mushroom locations for Gothenburg
or something like that. So, we need a bunch of data that
we can put it into this texture and drive the environment. And this texture, or sorry, this tile is
one kilometer across, and that is usually about
the size of the projects we usually work on. And we are making 700 of these
tiles to make up this grid. So it is a huge
undertaking in scale, but also in level of detail we are hoping to achieve
on this project. So, to finish this off,
hopefully we have a lot of Unreal interest in people here. I just want to highlight this
feature called Call in Editor. Since you are working
with all this data, it kind of puts a lot
of pressure on our workflow. So, we can’t really work
in editor in a normal fashion. We can’t click things
and move things around, because it is too much. So, Call in Editor allows you
to run Blueprints inside editor when even starting the game.
So, you get these buttons. So, if we want to import
thousands of buildings and place them in the world, we can just press a button
and it does it for us. Very useful if you work
with large data sets. And the funny thing is you
can make a button that presses
these buttons for you. And then you can make a button
that presses those buttons, and it is super cool. And then you can end up
with one big button in the end, and that is, everybody
who has worked with data and stuff like this knows
that it won’t really work, but it is at least a vision,
right, to have one press of a button, and then it generates
all the data, puts it together, names it,
catalogs it, and updates it overnight. And we want to finish
by showing you a quick look
at our progress so far, and thank you
for listening to us, and I hope you enjoyed it. [APPLAUSE]