Immersive flight simulation with photorealistic 3D tiles

In collaboration with Cesium, the Google Maps team have launched an experimental release of an API that allows artists and developers to make use of the 3D assets commonly seen within Google Maps and Google Earth. This dataset provides a vast quantity of 3D meshes of the real-world, all fully textured with accurate imagery.

Integration is incredibly straightforward and in no time at all, I was able to adapt my realtime flight simulator to make use of the new content.

Flying a Boeing V-22 Osprey low over the streets of Paris, in a flight simulator built with photorealistic 3D tiles from Cesium and the Google Maps Platform.

Flying a Boeing V-22 Osprey low over the streets of Paris, this video highlights some of the gorgeous buildings, visual landmarks, and parkland available. The content streams at around 60fps on a reasonably modern computer, providing a more immersive experience than previously possible.

Sweet valleys from up high

The photorealistic tile set, in combination with the Cesium plugin and Unreal Engine, makes for a dynamite combination of tools that create a stunning experience from a relatively simple setup.

It is notable that all of this was created using off-the-shelf assets, with minimal modification. I have already spent days and days simply flying around exploring the world.

Integrating the new experimental release of photorealistic 3D tiles from Cesium and the Google Maps Platform. Built with Unreal Engine, using JSBSim's flight dynamics model.

This video of a commercial airliner flying over the city of London and out towards the coast of the UK captures just a tiny fraction of what is possible.

Designed to be used in next-generation visualisation use cases, this mammoth data drop covers more than 2,500 cities across 49 countries. Crucially, during the experimental phase Google have said there's no cost to using the API, so now is a great time to dive in a have a look around.

Up close and personal

The main limitation of this product is that the data appears to be primarily captured by lidar and other sensors. This means the Google Maps team must have visited a place in order for it to be modelled. If an area isn't accessible, or the team haven't been to a location yet, there simply isn't any data.

This differs from the approach taken by Microsoft's Flight Simulator team, who have used machine learning to interpret what a place looks like from an aerial 2D image. This allows them to model vastly more places on Earth, albeit with less accuracy.

Another drawback is the amount of bandwidth required to stream this amount of material. A fast connection is necessary to obtain the smoothest experience.

I have not managed to get remotely close to the low latency demos that Google have been showing off at their recent Google I/O conference. This could relate to the experimental part of this release and hopefully things will improve, but using my connection it does take a minute or so for a scene to fully resolve at least for now.

Chocks away

If you're interested in creating your own flight simulator, there are many excellent resources to get you up and running. Some previous experience with Unreal Engine is essential, but beyond that, the following tutorials cover all the steps required to begin flying.

These additional free (at least at the time of publication) assets can help customise your airborne experience.

The progress of technology never ceases to amaze me. At the start of 2022 it seemed impossible for a single developer to create their own flight simulator. Then the Antoinette Project launched, and I could and did build one.

Back then of course, all the buildings were a single colour, their placement derived from OpenStreetMap. It again seemed impossible for a single person to realistically texture buildings on a global scale. And yet here we are not even halfway through 2023, and with a very small amount of effort I've pulled together everything you see in these videos.

Anybody can build this. And anybody can add value of top of it, just as I recently did with this ILS localiser visualisation.

I cannot wait to see what's new in simulation twelve months from now.