Home Defence, Training & Simulation How LuxCarta use AI to generate accurate, real-time simulations of anywhere on Earth

How LuxCarta use AI to generate accurate, real-time simulations of anywhere on Earth

How LuxCarta use AI to generate accurate, real-time simulations of anywhere on Earth

Building accurate and up-to-date environments has always been a challenge for military training simulations. Out of date environments are much less useful for training purposes, but keeping them updated takes a lot of time and money.

LuxCarta may be a relative newcomer to the military training and simulation (T&S) space, but they’ve been doing digital mapping for over 30 years. They’ve recently started using AI to build virtual models of any environment on earth faster than ever before. They’re now making these models available to T&S companies.

Novatech and LuxCarta often work together to offer complete solutions to military T&S customers (we work with LuxCarta to ensure that our image generators run perfectly with their software). 

Novatech caught up with Colin Urwin, Head of Marketing at LuxCarta, to find out how their background in telecommunications has helped them move into T&S and to hear about their new product, BrightEarth.

How did LuxCarta get started in Training & Simulation?

LuxCarta and 5G Network Planning

Most of what we do is related to satellite imagery,” says Colin. Through a process called photogrammetry, where two satellites from different places take a picture of the same spot on earth, you can determine the elevation using basic geometry. This allows us to generate 3D products from raw 2D source satellite imagery.

LuxCarta have spent years helping telco providers to map environments and plan their 2G, 3G, 4G networks. But the demands of planning 5G networks required them to find new ways to map virtual environments at high-speed.

Planning 5G networks requires you to understand your environment like never before. Trees, cars, vegetation - if these things get between the transmitter and receiver they can obstruct the signal. So the networks need accurate 3D maps for their 5G planning, which means demand for 3D mapping has increased exponentially.

This increased demand led LuxCarta to find new ways of building maps that were less reliant on manual work.

We were getting so much demand for 3D mapping that our old production methods just wouldn't work. So we had to explore using AI because having someone look at a satellite image and then build it manually wouldn’t have hit our deadlines. That’s how BrightEarth came about.

BrightEarth Titan Vanguard Terrain Generation

How will BrightEarth help T&S providers?

Earlier this summer, LuxCarta launched its BrightEarth suite of global imagery products derived using AI-enhanced production techniques. The new product line includes a 10m imagery mosaic and land use/land cover, along with population maps and building/tree extractor. BrightEarth will give training providers dynamically updated virtual maps of anywhere on Earth.

Colin says that the team’s goal is “Being able to go anywhere in the world and look at an image of the Earth that has been created. This product will get ingested into a lot of the simulators to create virtual worlds.

The models are generated using imagery from Sentinel-2 satellites and are then populated with cities, oceans, coastlines, farmlands and forests that are generated using AI. There will also be population maps, based on census data to approximate population size and density and how it changes at different times of the day.

“The big advantage of this is access to real-time or recent imagery. Something that we hear is, for instance, a client builds a simulator for a helicopter or a jet but the terrains and the maps aren’t updated. Or different parts of an organisation have different maps because they’re siloed. What we're building with Bright Earth will be something that someone could plug into and have access to the latest maps.

The importance of geospecific data

At a high level, there are two types of mapping data, geospecific and geotypical. Geospecific is where the maps correspond closely with terrain in the real world. Geotypical, as you might imagine, is when it’s a close fit or entirely fictional.

What we bring to the table is very much geospecific data. Because that's what we generate in the telco space. There isn’t much interest in the geotypical data for RF planning, they want to know exactly what the environment is like and that's what we give them.

This has particular benefits for training simulations where trainees are preparing for deployment in specific locations around the world.

“For operational stuff or mission planning, T&S providers will want the best information they can get. ‘This is the environment and these buildings are exactly where they're supposed to be.’ That’s something that T&S people love, the ground reality. The buildings and the trees being where they’re supposed to be.

One of the challenges of building maps of different locations around the world is that the terrain differs massively wherever you go. Algorithms need to be retrained in order to realistically populate maps from different locations.

The dataset that you build for Brazil is dramatically different from something you build in North Africa, which is dramatically different from something you build in Malaysia, which is dramatically different from something in Western Europe. Luckily we have manually produced datasets from all these parts of the world that we can use to train our algorithm to replicate the environment.

Thanks to Colin for taking the time to catch up. You can learn more about the product on their website.

Images copyright TitanIM

Novatech supply custom, purpose-built hardware to some of the biggest names in the Security, Aviation, Defence and Marine industries.

If you have a project you'd like to discuss, please contact our dedicated Simulation and Training team using the form below, or call us on 02392 322500.


Posted in Defence, Training & Simulation

Author -

Published on 30 Oct 2020

Last updated on 30 Oct 2020

Recent posts