Upcoming Events

Weekly BMW races on Simracing.GP Other regular AC events on Simracing.GP Weekly GT3 Sprint Races on Simracing.GP Rookie friendly WTCR sereis Weekly rFactor 2 events

Could this be the right time to start an Open Source Sim Racing project?

Steve Worrell

Staff
Premium
Messages
1,051
Points
2,304
In addition to being an open "kernel" for sim-racing.. additional features like loose surface physics and freeroam/traffic would make it the ultimate sim-driving platform. :D
Ok now we're talking. Bram and I have nerded out a few times about how cool it would be if all the sims talked to each other. So you could race on the track with say, acc, then transport your car in Truck Simulator, then fly your cargo in flight sim etc.

Ok, I know, i'm getting wayyy ahead of myself here but just wanted to share that thought.

But on a more race-sim focused feature, an idea I have had for a long time is about damage to cars. It seems every time I ask about damage the answer I get is because of licensing. So my solution to that would be, ok, let's make some fictional cars which demonstrate the power of the damage model and then let the community create mods. A bit like how Konami create their football game with fictional teams, and a month after release the community has made all the kits and players as a mod.

Forgive me if I have misunderstood the complexity of damage models in sims but I'm desperate for an experience where you really have to look after your car. I've been banging on about this for years - For more information on what I mean, check this article I wrote in 2017, The most realistic sim I have ever played came free with a pack of cereal...
 
Last edited:
Messages
34
Points
152
Forgive me if I have misunderstood the complexity of damage models in sims but I'm desperate for an experience where you really have to look after your car. I've been banging on about this for years - For more information on what I mean, check this article I wrote in 2017, The most realistic sim I have ever played came free with a pack of cereal...
While I agree on the more realistic damage model, such a thing would have to come with top notch AI, as the divebombing, unaware, hitting you mid corner stuff that is commonly found in many sims would end up screwing over your race 9 times out of 10. And even if you can learn to avoid the AI in their trouble spots, I can see many cases where less than half the grid finishes because the AI crashes itself out doing something dumb. Especially true in multiclass races where faster cars can lap several seconds faster than the slower ones, and usually end up packing the rear of a slower class car in the process.

Then again, I would love to see AI improvements across the board.
 
Last edited:
Messages
61
Points
62
I wouldn´t put much faith in ai; seems like the only sim that has a consistently good (decent?) ai is iracing and if the amount of income that game has is what it takes to get it right, then I´d rather see dev time spent elsewhere.
Otherwise, the amount of options that rf2 has in that department gives pretty good results when you put in the effort to tweak it, despite the core ai being broken.
 
Messages
34
Points
152
I wouldn´t put much faith in ai; seems like the only sim that has a consistently good (decent?) ai is iracing and if the amount of income that game has is what it takes to get it right, then I´d rather see dev time spent elsewhere.
Otherwise, the amount of options that rf2 has in that department gives pretty good results when you put in the effort to tweak it, despite the core ai being broken.
Funny you say that, because atleast on Ovals from what most people have gathered, the behavior is basically the same as/if not very similar to a well dialed in NR2003 track. The road racing side is probably MUCH better, but mainly because iRacing has cars are actually built for road racing rather than in NR where you try and shove a touring car onto Stock Car physics.

Don't know if that says something about iRacing's AI progression, or how right Papyrus got it back in the day that smoothing it over and providing proper cars for the discipline is enough to put it in the same class as other much more modern systems.
 
Last edited:
Messages
10
Points
36
While I agree on the more realistic damage model, such a thing would have to come with top notch AI, as the divebombing, unaware, hitting you mid corner stuff that is commonly found in many sims would end up screwing over your race 9 times out of 10. And even if you can learn to avoid the AI in their trouble spots, I can see many cases where less than half the grid finishes because the AI crashes itself out doing something dumb. Especially true in multiclass races where faster cars can lap several seconds faster than the slower ones, and usually end up packing the rear of a slower class car in the process.

Then again, I would love to see AI improvements across the board.

That‘s where I‘d like to throw machine learning into the mix. Let the AI simulate a couple thousands of laps in various scenarios and it would improve over time. Players could opt-in to share their session data to help feed the learning algorithms.
 
Last edited:
Messages
402
Points
462
That‘s where I‘d like to throw machine learning into the mix. Let the AI simulate a couple thousands of laps in various scenarios and it would improve over time. Players could opt-in to share their session data to help feed the learning algorithms.
There's more work involved than just "throwing machine learning in there", but yes, I agree that such an open source sim could be a great development environment for racing AI algorithms that could at some point yield very nice results. You can approach this in a number of ways, and I'd be very interested in seeing some machine learning reseach / development being performed here. I might also be interested in getting involved, if my time allows it (I work as a researcher in AI/ML).

On that note, what happened to the AI in that Moto GP game a few years back? I saw a lot of marketing hype around it when it was announced, but never heard any of the implementation details, nor did I ever hear about that game again...
 
Messages
6
Points
16
An open source framework for motorsport simulation could be an interesting idea to share and evolve the knowledge in how to build motorsports simulations and be a focus for experiments and innovations (e.g. smarter AI). We shouldn't have to reinvent the (racing) wheel when a new generation of games are made.

I think one issue will be to attract and keep a critical mass of contributing coders. If too few people are making material contributions, that small group eventually get tired and the project stalls.

In the case of a motorsport simulation I'm guessing there is a narrow subset of people who have the combined skills in coding, real-time software simulations and motorsport engineering. So it might help if the project is also a focus for information that explains how such systems work and how physical properties of cars are modeled. Many people have coding skills but the application domain is the niche that few people understand.
 
Messages
72
Points
52
Somehow missed this til now – whoops! Super exciting idea. Especially when I see who's talking about taking it on! Wow. So cool! :)

And what Stefano emphasized is his initial post is exactly what made me start a thread or two about this earlier this year – if Ilja/Peter Boese/etc can do what they did with AC without the source code (although, to be fair, AC is quite open), imagine what people like that could do, if they so desired, if you make a good sim 'sandbox' to build on!

Perhaps a result to hope for is what the ISI engine (or whatever it was called) was in last-gen sims, underlying rF1, GTR1 and 2, a couple EA games I think, GSCE, and AMS1. Except now, instead of developers having to pay ISI for licenses to use the engine or get the source code, the engine would just be available to use as desired – including in commercial projects, which would require MIT license (as Stefano said) or alternatively a BSD license.

Another benefit to something like this is to ensure the continued existence of a high-quality, current platform supporting modding and community content as a first-class citizen.

I personally would be very happy to sponsor/donate money and get involved with this project and contribute some of my spare time! Although I have few skills to offer though, haha, so I'd be limited to non-technical stuff like research, documentation, or bug triaging.

Quick sidebar – small, "grunt work" like bug triaging is super important for the success of open source software (see for example this link about KDE, a popular desktop for GNU Linux). The point? Pretty much anyone in the community here at RD can get involved with an open source software project like this, if it were to become a reality. Not just modders and coding ninjas.

EDIT: When I was interested in the open source sim idea the first time, the research I did revealed a lack of open source car physics engines (especially for high-speed) intended for real-time driving (not just calculations). So another benefit of a permissively-licensed open-source sim engine would be to remedy this gap. It could even be the basis of non-game applications – be it a self-driving car sim or a basis for racing teams' custom physics.
 
Last edited:
Top