Flight simulation in 15 years? (Horizon scanning)

don’t mind me, just jacking in via USB-H and loading flight sim directly to my cerebellum…

2 Likes

you probably aren’t far off

Hello,

If I may offer a slight correction to your list, photogrammetry comes from aircraft only a few thousand feet above ground level, not from satellites.

Time to take my moderator hat off for a few minutes, and add the disclaimer that I don’t work for Microsoft or Asobo, and am only coming at this as an opinion from a professional software developer who does not develop video games for a living.

Fifteen years is a long horizon to predict. There may likely be a paradigm changer that only a few of us could foresee today. I think a lot of us forget that the paradigm-changing “technology” that we’re living through today is the cloud. It’s the defining thing that sets MSFS apart from its competition.

AI

The safe paradigm-changer to predict for 15 years from now would be AI. While it does a lot for us today, it has the potential to be inconceivably different in 15 years. Think ATC speech recognition and natural-language interaction. Imagine each AI traffic aircraft with one or two thinking AI pilots that may not always hear the AI ATC correctly and make occasional mistakes, or that may assess situations on their own aircraft that affect their interaction with the outside world. You may opt to have your own AI first officer or co-pilot who may follow company procedures, but also has their own opinions on certain things. I don’t think that the barrier will be technology, here, but instead, the team’s resources and willingness to develop this. I think there’s room for third parties to grow, here.

Scenery

As far as scenery and AI goes, I’m thinking that in 15 years, we might lose this distinction between photogrammetry and autogen scenery. Currently, we have static photogrammetry data that gets heavily processed for performance, and then manually cleaned up by a team. I think in the near-term, we will see big improvements in photogrammetry, as processing improves and the system can take, say, a glass building and infer that it’s mostly glass and give it the properties of glass. But in the longer term, I can also imagine a world where we don’t have procedural buildings, and AI simply processes the world with whatever data it has and generates highly-detailed buildings and scenery everywhere. Imagine taking an overhead picture of trees in the spring or fall seasons and inferring the various types of flora in the area.

Weather

I think weather will be out-of-this-world great in 15 years. Weather is incredibly complex, and to do it right today requires supercomputers. Fifteen years is not going to change that, and the majority of that processing would still have to be done in the cloud. But I think that the fidelity available to us will be spectacular. I think that at that point, the bottlenecks will be the accuracy of the stations reading the data on the ground and the models that meteorology can provide us.

Aerodynamics and Physics

I think physics stands something to gain, but maybe not by the same leaps and bounds as weather. I don’t remember how many points along the aircraft we sample today – a thousand? Would it help to double that to two thousand? Maybe? (I really don’t know.) But at some point, we will hit diminishing returns. I don’t know when that will be, but I don’t think it will take 15 years to hit that limit. I think that most of the gains to physics will happen in the next few years as the team replaces legacy FSX code.

Third-Party Software

From a third-party standpoint, I imagine that there will be a study-level third party aircraft for most popular aircraft. Assuming a healthy market, I’d love to think that nearly every class B airport in the world will have a Drzewiecki/FlyTampa/Pyreegue-quality airport available, along with most popular class C’s. Though I do wonder if advances in AI might one day push some of these scenery/airport companies out. I don’t know, and it’s not something I want to think about because we need them so badly in today’s world. I do wonder if third-party mesh scenery might disappear as Microsoft goes around the world with their world updates: The niche I could see would be cases where Microsoft doesn’t want to pay for some expensive dataset for some part of the world, but a third-party developer might be willing to take that risk.

Third-Party Hardware

I’d love to say that this would continue at a high pace as more and more people own Flight Simulator (and dare I say, that other simulator that starts with an X), but I don’t know. This is a niche market and will continue to be a niche market. I do think that some of the barriers for the average simmer to create the rig of their dreams will continue to lower. I think there will be more options to create your own airliner simulator at home, but the market will always be constrained by people’s home space and significant others who don’t want you transforming your basement into an A350 cockpit. :laughing:

I’d love to see third-party hardware manufacturers come up with modular systems that allow you to build a home cockpit, piece by piece. You’re seeing a little bit of that now: At the consumer end, Virtual Fly has a new modular series of cockpit switches called Switcho, and you can start with as few or as many as you want and build on it. Honeycomb was working on a gaming chair (I don’t know if they still are) that would pair with their future Tango Foxtrot sidestick and throttle. At the highest of high-end (this is really for commercial customers), Brunner has a modular platform that allows you to integrate many of their components into one place, and Precision Flight Controls also has a modular enclosed system, as well.

The Platform

What will happen to Microsoft and Asobo? Again, I will stress that I don’t represent the developers, so none of this is based on any internal or privileged knowledge, but merely my own guess as an enthusiast and layman. I think that as time goes on, we are likely to see a slowdown in sales, as there is only a finite number of people in the world interested in aviation, and the market will reach saturation at some point. I think xCloud will certainly help reach gamers in the developing world, and that growth would continue steadily for many years. (I actually do wonder if xCloud is the future for all gaming some day.)

That is not to say that there will not be new sales, but I just can’t see it keeping the same pace as it has for the past two years. But I’m thinking that even as sales of the sim slow, the Marketplace will continue to generate revenue and sustain operations. I also think that Microsoft would be foolish to keep this digital twin solely for Flight Simulator. Not only could they reboot a franchise like Train Simulator or Combat Simulator; I’m sure that they could spawn racing games or FPS games or other games that rely on a representation of the real world. Or they could branch out and make a historical world with the same weather and physics engine. (In fact, why reinvent weather and physics in every game you produce?) I’m sure that there are business applications way beyond aviation, too.

Conclusion

Again, these are my own thoughts and predictions. Don’t beat me up if you disagree; I’m coming at this as an enthusiast, not as an industry insider.

I said earlier that there may be a paradigm changer that many of us don’t even know about today. Fifteen years ago, the cloud existed, but it was not nearly as ubiquitous as it is today. All of my predictions are grounded on what I know exists right now, but I’d like to think that there’s another disruptor around the corner, waiting for us. If I could predict what the next disruptor would be, I’d probably be a lot wealthier than I am. :smile:

6 Likes

I can definitely see weather improving over the next 15 years, like you said. It is not just the processing power that has held weather back, but also the basic science. We have only had high quality wide area weather observation for 30-40 years now, so there hasn’t even been that much reliable data to attempt to build models and sims off of.

On physics, I suspect we’re probably at the limit of what we will see from aerodynamic modeling. I would suspect the next areas of aircraft modeling will probably be on the mechanical and thermodynamics sides of things. One of the really fascinating aspect of the mil-sims are what happens when the plane is pushed to or past its limits. Over-G or over-speed and parts will bend, flutter and sheer off. Land with too high a sink rate and you’ll have a main gear wheel following you down the run way.

One of the most memorable pieces of flying I recall reading was a description of a flight in Jane’s F/A-18E Super Hornet, where the plane had what seemed to be a minor engine issue, that, over the course of a 20m flight back, spiraled into a full on emergency, with multiple major systems failures and both engines ready to fail outright.

I suspect we will start to see that sort of modeling start to make its way into the civil aviation simulation world too, especially as programmers start to come to grips with how to multithread flight sims.

I suspect the modeling side will also end up leveraging AI for ingesting research data into sims. Right now doing those mechanical models are slow and expensive because you basically need a MechE to convert the drawings into models, but I could see AI being used to do a lot of the legwork and even being used to make reasonable guesses to drastically reduce the level of effort to do that.

On peripherals, I suspect we will continue to see more small scale entrents to the market. While it is a niche market, the barrier to entry is far lower than it used to be. Between things like 3D printing, readily available programmable microcontrolers and FPGA, and standardized inrerfaces, it does not take a large capital investment to start making peripherals. It just takes skill and wanting to do it.

Force feedback will probably take a bit longer. Because of the patent issues over the last couple of decades, I don’t really think it has the standardized interface it needs to go wide yet. Once that happens, I think availability will increase dramatically, but I’m really not sure what it would take to kick that off.

VR, I have no idea where it will really be in 15 years. We will definitely have the hardware to run it sufficiently by then. But it’s such a big change from the conventional keyboard and monitor system that it is hard to predict what form and how ubiquitous it will be. I think it will definitely still be a significant player in Flight simulation; the depth perception and vision control is just to essential to certain classes of flying for it not to be, but that’s all I could really guess about it.

We will probably also see raytracing fully developed as a system, and common enough that we’ll likely even be seeing it used for non-graphics things like AI path finding and radar.

2 Likes

Agreed! In my statement, I wrote, “I think that at that point, the bottlenecks will be the accuracy of the stations reading the data on the ground and the models that meteorology can provide us.” When I mentioned models, I was referring to the models produced by our current understanding of meteorology today. Hopefully as technology improves, we are able to further research in meteorology, as well!

Totally agreed on aerodynamics, and you have a terrific point about thermodynamics that I didn’t even consider!

Excellent thoughts!

In 15 years time, I think that flight simulation in particular (and maybe gaming in general):

  • could be fully streamed and be cloud based

  • VR will be mainstream and PC monitors will be the preserve of the minority

  • monthly/quarterly/annual subscriptions will finally have become generally accepted as a fair way of paying for regular sim updates and to fund server costs

  • audio will be much more immersive and realistic than it is now

  • haptics via flight sim/gaming kit (e.g. flight yokes/sticks, gaming chairs etc) will have grown in importance and will be much more common/better than now

  • scenery will be just about photo real

  • third party developers will still be here to give an even better sim experience than that given by the vanilla sim

I think that we will also still be complaining that flight sims are still not yet advanced enough to give a realistic impression of real life flying.

How did you dare to correct me!!?? :slight_smile:

I totally agree with all you said. I think NLP, one of the most advanced parts of AI, will play a massive role in our future life. Flight lessons, copilot cooperation, ATC are likely to be completely managed through AI. Actually, I think the interaction with machines will.

Your point about autogen no longer existing as we know it, is something I totally agree upon. I was making exactly the same point before. I want to stretch by saying that sensing technology, being it LiDAR, visible spectrum imaging, Vis/NIR satellite, or even emerging technologies such as satallite radars , will create a completely different space of data. AI will not just be able to build shapes but, for example position of water thanks to its different radar reflection.

I think it won’t take long before we get modular cockpits but the key will be VR plus haptic technology. Physical hardware will always be limited by space and physical similarities of planes. Much easier and functional to get portable devices!

I am not sure I understood completely about your point about cloud computing. While in general I would 100% agree, not sure why in a rendering exercise it would be usable?

Btw, thanks so much for your contribution, so interesting to see your views. I did lot of research on AI (I actually manage lot of research atm) and multimodality (a 10y old PhD) but I often struggle to understand what people think the advances will be. Thanks!!

1 Like

Your comment is very stimulating. What do you have in mind with thermodynamics? Super/hypersonic physics or better engine representation?

While for military planes, mechanical distortion or failure is very important, not sure about a civil simulator. Not sure about the use of Machine Learning for that, just because you would struggle to find a decent dataset to train it on.

I feel that the aerodynamic model will be little improved in the future too, but out of a different reason IMO. You won’t get very little benefit from further improvements while you are sat on a chair in front of a screen.

I 100% agree with everything else! Please don’t take any of my comments as challenges, I just thought your post very very interesting!!!

1 Like

Hahaha! (Or should I say, “Mwahaha!”)

Excellent points on all of those various technologies. Satellite radar is an interesting one, and I look forward to the day when the ground layer isn’t speckled with occasional clouds.

Also, an interesting thought about haptic technology! Right now, I have a single Buttkicker under my chair, but this is definitely a less-explored area at the moment.

Which line was it about cloud computing that you’re talking about?

I only have a Masters in computer science, so my knowledge of AI is :pinching_hand: compared to yours. But I definitely think there is a lot of unexplored potential there!

1 Like

Very sorry, I mixed up your comment with a comment above. I withdraw the question about cloud computing hahahah

1 Like

I have been trying, for Xplane, to use mesh data to correct photorealistic imagery. I started developing algorithms that aligns ridges in the imagery with ridges in the DEM. It’s to try to avoid ortho where the image has the ridge represented few metres away from the actual point in the mesh. I am also trying to substitute stretched images in high slope areas with pseudodata. So far I have consistently failed :slight_smile:

I could see this sim being a testing for AI assisted real world ATC.

Simmers are definitely going to throw the unexpected curve-balls AI learning would need.
A lot of this is possible today, with the right hooks into the sim being available. 3rd party developers simply need the access (and of course the market) to make it happen.

It’s the same reason people are asking to open access to aircraft config files, the weather engine, the ATC engine, and lots of other parts of the sim. By allowing people to tinker a bit more, surprising solutions and advances will come about.

1 Like

On the thermo side, yes I’m mostly thinking more detailed and physics driven engine and systems operation and failure modes, particularly for piston engine operation.

And I am expecting the structural modeling to mostly be developed in the milsim side first and trickle into the civilian market. Still, I have seen that wing flex is not a trivial part of turbo prop and modern jet airlines behavior.

AI, I’m expecting to be used more in the development side. It takes a lot of time and effort to take an older set of paper drawings and turn them into a simulator model. That goes double for aircraft that the drawings may not even exist for, so I’d expect us to see AI tools used very heavily to do the initial ingestion of raw aircraft information into initial models.

I can also see AI tools being very useful for converting true CAD models into good fidelity simulator models. Actually, I think we may see that first. Producing a low resolution mechanical mesh that closely matches the behavior of a true resolution Finite Element CAD model seems like exactly the sort of thing Machine Learning would excel at.

It may even be possible to do that from the Original Manufacturer’s CAD models without revealing trade secrets. Though I imagine that involves more than a few lawyers getting their wings, or at least boat payments for a few years…

It’s going to be some really interesting stuff.

AI + Quantum Computing will take us to places we can not even imagine.

2037…

DX16 released and now no need for SU10 and DX12 to be released. Apparently, it is still coming “soon” anyway.

The franchise of Microsofts wildly successful Opening Door simulator is to encompass some elements of flight simulation for added immersion.

The Flight Simulator forum surpasses Wikipedia in terms of sheer size. Now takes two hours to open, even with a 34900K and 512GB of DDR14 ram.

I feel sure that some wag out there might like to add more, so feel free.

Thanks, I really appreciate the time you put into thinking about such advances. Well, I certainly agree with the ethos, not sure if I agree with the topics. Not sure machine learning would improve significantly on the existing engine calibration curves. Advances in 3d scanning will likely help any form of computer graphics, hence I think you are right in thinking of model development aids. Solid mechanics is a quite well understood problem, we may get good statistical models (again, hard to train such model on actual datasets) showing failures, but I really think that deterministic world is not over, at all :slight_smile:

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.