As you are directly involved with the physics, would I be wrong in saying that you are also limited by the cost of research and hardware as it’s something you did not mention? Up until the introduction of Ryzen, and competition with Intel most main stream processors were 4/8 thread max. With the complex calculations and limitations of games/sims only using a couple of cores and the calculation and rest of the engine taking up resource is or was this a restriction on the physics. With what you have said, if you only get limited data from the tyre manufacturer, could you see franchises like turn 10 (Forza) and Polyphony Digital (GT) Start to dominate the sim market, because they have much deeper pockets, and they will have comparable hardware to pc’s shorty.
Good question! Here is an essay.... I feel there is this 'urge' in people (engineers) to make things more complicated when the processing power allows. I also feel that customers get a good 'feeling' knowing whatever they're playing is incredibly complex. What in my opinion happens way quicker than most engineers would like to admit, is that you're drowning in a sea of numbers. You simply can't control a system with a lot of variables. And the goal of a game is not to be some PHD tire model project. The goal of a sim / game is to deliver good content that is consistent from car to car, with improvements from beta tester / driver feedback being 'easy' to roll out over the entire car lineup. Too simple is bad. If there is just one 'grip' value, then it is easy to adjust but you will be limited to what you can change in the 'feel' of the cars. Too complicated is bad. If you get feedback but don't know how to adjust your model for this car, let alone the other 50 cars, you're not going to get anywhere. Since I've made so many mistakes over the decade I've been trying to understand physics, I am very cautious about adding complexity. People who feel differently about this COULD be right. But if I was twice as smart, which is probably the upper limit of what 'physics guys' can be, then I still wouldn't be able to control a highly complex system. Perhaps Dave Kaemmer and Terence are more than twice as smart, but they don't tune the cars in the game. You see I've learned to be extremely wary of complexity. People overestimate themselves all the time. No matter if its blind wine taste testing, blind audio equipment listening, or blind sim car setup testing, we expect to taste / hear / feel a difference, until it is a blind test. Then we come up with excuses rather than admitting to the placebo effect. It is probably natural human behavior to be over optimistic about our ability. This can really hurt race sim development if you don't constantly ask yourself 1) Do we really need this complexity 2) does it really add something we can feel / measure / validate with reality 3) does its complexity not hurt us in our development time and keeping the sim up to date If you don't constantly ask yourself this, and you get carried away by our human tendency to say 'Hold my beer' and do something complex, I feel you get too complicated models that are too hard to work with because you've lost sight of your own limitation and the reality of producing and maintaining sim content. The most important thing to add here is that simple is NOT simple to 95% of people! The code is still complex, the models still have many variables and it is very difficult for me to see through the AMS tire model and get the most out of it. So it is not simple, it is not arcade. Processing power or money to test more tires is not a factor in my opinion. The weak link is the human, not the machine.
@Niels Heusinkveld Thank you for taking the time to reply and clearly explain and expand on your video. I completely understand what you are saying now and understand the limiting factor 👍🏻
@@NielsHeusinkveld Well said. I work in IT designing and maintaining systems and networks, and I am constantly asking those 3 questions. I think it is vitally important across so many fields, and is so often overlooked. Is a more complex system more or less likely to break? The complexity may solve one or more problems, but adds in more elements that "can" break. And if they do break, as they are more complex are they even harder to fix? So in the end are we worse off?
@@NielsHeusinkveld Words of wisdom. It reminds me of the probably apocryphal story of NASA pending a ton of cash designing a pen to work in space where as the Soviets used a pencil !
Actually Niels - Douglas Adams may actually have the perfect quote for you here: “The ships hung in the sky in much the same way that bricks don't.” Fascinating video as always, thank you.
I believe Douglas also used to fix problems by completely ignoring them, something I frequently copy. If you haven't seen his video about his trip to some tropical island for a nature documentary, its gold: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-_ZG8HBuDjgc.html
This video is hands down the most informative I've ever seen regarding sim tire models. I hope you keep giving us your input on this weird hobby of ours. Cheers!
Wow Niels, you got all sim community involved. Tyre model is one of key elements that defines the joy of driving a sim. I would love to watch both you and @Aris.Drives discussing the subject on a live stream...
As someone who works on mechanistic modelling for a living, you cannot get around the lack of data by using mechanistic models vs. empirical models. You still need to validate mechanistic models against experimental data to understand if the simplifying assumptions you made affect the accuracy too much. iRacing's modus operandi seems to be to just declare that since it's a mechanistic model that means it must be more realistic than other models and to ignore all the anecdotal data that says otherwise (like pro-drivers complaining).
azarisLP there are books, Segers is an example, that shows real x iracing telemetry data. The flaws in low to medium speed cornering are huge. Despite having a huge partnership and access to teams, they don’t seem to do the QTG properly
There seems to be some confusion over criticism of the iRacing tyre model - it's not forgiving - and the pros and cons of physical and empirical tyre models. There are empirical tyre models which are unforgiving and physical tyre models which are forgiving.
As a software engineer, race car driver (not pro, but i race pro's ;)), and someone who's been trying to get into more sim racing the last few months, thank you very much for the videos. Lots of cool insight into what's going on and why I'm finding it difficult to enjoy. Cheers!
Interesting Niels, but from what i've seen with the rf2 model, it has the physical model, but it also has many (more than 60) modifying parameters used to tweak the behaviour of the physical model (in the 'realtime' section of the TGM) . So in reality it's a hybrid physical+empirical model. But in the end, if a physical model produces certain peak slip angles and so on, you can't really use fudge factors to get correct behaviour, it just becomes errors upon errors
Very good point, I know it has some room for tweaking, but I don't know the extend to which you can. I do know that the RF2 physics guys aren't as 'free' to tweak as they sometimes want though.
I think real-time section in rF2 allows tuning things like wear and grip levels so pretty fundamental things:). I’ve read GP3 modder worked for months with Michael Borda. It took few iteration of creating tires, comparing with real telemetry , tuning values and comparing again etc. Real drivers feedback was also involved. Same tires will be used in upcoming F3 mod. But that approach is of course an exception.
@@NielsHeusinkveld Do you know to what extent S397 have control over the core physics of rF 2? From what I've been able to read between the lines, the base physics of the game were locked off by ISI, and S397 have essentially just been a professional modding team, adding a ton of licensed content to fill out the game. At least, it's the only thing that explains the many glaring issues that have been allowed to remain in the game for so long.
Curious to know what your take is on BeamNG's tyre model. They seem to use a physical modelling approach for the entire car, not just the tyre, leading to some fairly astounding emergent properties such as natural chassis flex and force transference. blog.beamng.com/a-look-at-tire-development-in-beamng-part-1/
BeamNG is just super impressive as a whole. The handling does seem quite twitchy still but I can't really tell if that is because of poor models or poor data going in..
@@NielsHeusinkveld Well, PC2 is much more twitchy if you ask me... By the way, I suppose it also has an analytical model, so it should be quite difficult to work with too? Inspired by your other video I decided to give AMS2 a try to see how much more of an improvement it is compared to PC2, but... Braking and accelerating out of corner feel really weird to me, honestly. In braking I barely have any feel to what the front is doing, while pressing the loud pedal feels like I'm controlling the rear step out directly in a linear fashion doing so. Caterhams are especially guilty of that while RR cars feel more or less like I would expect a car to feel. I have to admit that without these issues driving feels very much immersive though, but I'm seriously disappointed with braking, which is my biggest joy usually (trailbraking in particular). Is there a more drift-ish car in the works, by the way? As for BeamNG, sure, not all the cars are perfect, but if you put the LS swap mod into the drifting version of the 200BX car, it becomes a complete gem. Also, muscle cars are very nice, as is the recent "WRC" type car.
@@NielsHeusinkveld Glad you feel that way. Their content is definitely very hit and miss, but the ambition behind what they're doing is mind-boggling. If you get one of the better cars, it's like magic. It renders weight transfer unlike anything else. It's essentially the best road car and drift simulator on the market, even in its strange quasi-sandbox state.
@@NielsHeusinkveld chassis should be physical for perfect aero and tires should be empirical pacejka for perfect traction and grip and slip. Beamng handles weird because it has infact a physical tire model which imo is the wrong approach because of what @Niels Heusinkveld explains in the vid above.
in my Engineering School, some engineer from DAMS once came up and he told us that their simulator (worth around 1M€) doesn't feature any tire model for temp and wear (so always in ideal conditions) because it was too complex to make an actually realistic one. So I guess no matter what video games developers try, they will never match something realistic enough, considering they have a lot more time/budget constraints than professional simulators
Great video. I can’t see paying a monthly bill when so many great racing games are out there to buy. But I’m glad you pointed this out on last video. Enjoyed watching 👏👏👏👍
The Portuguese Gamer the main reason is how much updates and new content is added. Alot of single purchase games disappear quickly, or get replaced. iRacing is a 12 year old game...
CalvinnnnG true. But when they get replaced I just buy the new one. Once. Maybe a dlc but that’s it. And I can play the older ones if I want. Because I own them. I think if they changed a monthly bill and all was included then it might be worth it. But to pay a monthly bill a then have to purchase cars and tracks on top is crazy to me. And if you stop paying monthly what happens to everything you bought. You can’t even use them. Unlike RaceRoom or r factor. Yes I may have to buy cars and tracks but I really own them. And they look fantastic in VR. Lots of online competition. And in RaceRoom most are free with prizes to win. I get that people that are in i racing love it. But it’s just not a good deal in my view. And I love sim racing. Have a complete simulator that I play on for hrs. But that’s just me. I can’t even see paying for cable when there is so much programming out there for free. Like I heard one RU-vidr say. For the cost of one year with cars and tracks. I could buy a new dd wheel . But I thank you for your comment.
Thanks for the quick course in dynamic tire simulation models. Good to learn from your insights. I wonder after ‘physical’ tire models are repeatedly adjusted to sim performance goals, the geometries and elements get further away from real tires. In such case, their ‘physicality’ loses the argument of reflecting reality.
Thanks for explaining that makes a lot of sense. This is way more complexed than we could ever have imagined, especially in iRacing's case since they have been getting lots of negative feedback regarding their tyre models
What I find interesting about iRacing's physical model is that there is still some of the old 'skaty' feeling that was there right back to GPL. As an approach the NTM aims to objectively represent a real tyre but it still totally has DK's stamp on it. As a comparison, and a way to explain, I always felt that the tyres in iRacing feel like you have hundreds of tiny ball bearings on your tyre - you don't really get the kind of chattering grab and slip feeling that happens 100s (or 1000s) of times a second on a tyre. I do get that feeling from Asseto Corsa and NKPro before it, although I'm unsure what model type Kunos employs. I find the supsension and chassis physics in iRacing more advanced though.
One of the most lasting lessons I learned from a 2 year engineering degree came from a meteorology teacher. He was very good at explaining things in simple terms freshmen could follow. There is no way I can say as simply as he did, but I'll try. Basically even with all of the collective knowledge of human beings we know very little for a fact. And individually most of us know nothing for a fact. Tires are very very hard to understand let alone model them. And I'm sure we have many scientific holes in our understanding of tires. We sure did when I studied them in the 70s, 80, and 90s.
Thanks Niels for sharing your knowledge with us so that we can comprehend the various aspects of the needs for accurate sim racing tyre modelling. Much appreciated.
Next: how does netcode differ between titles, specially how they fill in small gaps in data, do they just extrapolate or use something more sophisticated.
There can be things as simple as minor delays in positional data , right through to great burps which is where the predictive calculations come in to basically filter a cars position, but overlay the directional change and vertical acceleration and deceleration. Now if we then couple to this the varying level of grip due to temperature changes and the loads that the tyre is subjected to and the temperature calculations that must be done as a result of this compromise . The grip and wear and the rate of temperature rise IS related to the grip factor of the particular area within the sim, the traction change from Grass to loose gravel to quality tarmac . All of these things are difficult and time consuming to say the least .
@@mozzmann I know something about this and i highly doubt any of the titles use very complicated methods, it seems like it is mostly based on just simple extrapolation where the vehicle will follow the current vector. This works very well on straights but are awful at corner at high speeds... It still is true that you want to be on the inside if the other car is lagging even a bit since all errors are going to point to the outside of the corner.. I used to be against ghosting but now i think it is the only way to really solve eSports lag spikes. You disable vehicle-to-vehicle collisions first, then start looking at "where is this car suppose to be in this frame"..
@@rkan2 Yup, that is just how time goes, you can't rewind so it is going to be lossy. What i'm actually interested in is that is it just simple extrapolation based on previous frame or is there something that at least tries to make it plausible behavior. Simplest extrapolation means that while turning and losing packets, the vehicle will continue straight and stop turning.. It retains velocity but not rotation. F1 2019 has ghosting and AI control at some point when you lose connection and i hate to say it.. i think their way is the best...
They ran into this issue in Dirt Rally 2.0. The tire model degradation was so complicated that when in QA testing just before release the realized the soft tire had too much life to it and they wanted to reduce it but the people who designed the tire degradation model weren't around and they couldn't figure it out so I think they kept it. (The tire degradation had a lot of promise but it was too complicated to understand/have enough time to tweak it perfectly).
Very interesting video. It almost seems like a really good way to make a good feeling sim would be to just get as much pro driver feedback as possible and use the empirical method.
Thanks for the tech talk and all that...however..double thimbs up for the Douglas Adams documentary...absolute legend of a verbalist/author...genuine laugh out loud...cheers
Excellent explanation, Niels. I couldn't agree more. I am being involved in modeling physico-chemical systems since 1990, and computation has made the change. Models need calibration and validation and for that, you need extensive, reliable and accurate data sets under different coditions. As you say, data on tire behaviour is not always available for those purposes. It is very good that iRacing and rF2 make an effort to advance in this sense. I would encourage them to keep working. At the same time, it is perhaps good to review the scientific literature and search for simplified models, still with a physical base. Thanks for the explanation Niels!
I'm happy with the feel of say, what Polyphony has done in GT Sport, but I guess it's a combination of tyre and suspension physics combined....in the end, regardless of what sim or racing game we play, I'd like to think the majority of us accept the limits of not only all the other calculations a "sim" has to do, there will be compromises, what's important is we get the balance across all aspects that meet our expectations of racing cars, where I'd say 99% of gamers have never even raced a professional race car, so how can we really claim the tyres don't work correctly. I'm happy to let the devs do that work, and I'll support it. You know best!
Forgot to mention one factor that no doubt comes into play on the limited data - tire manufacturers not wanting to give away their trade secrets. I bet they have some of the data you guys would love to get your hands on, but they're fearful competitors could figure out their secrets by reverse engineering the data. Great video!
I really love those Videos. It helps to understand the problems the developer have. I believe by better understanding the issues, the community will be able to give more helpful feedback to the developer and hopefully speed the process up a little bit. It's also just hugely interesting to hear about that stuff. Hope to see more of such content.
I've noticed via modding experience that most of the time it seems like sim developers get almost nothing useful from manufacturers, even for suspension and other things which are easier to measure. Let alone tires!
Yup, they may not like to admit it but I am certain some big name cars in some big name sims were made with little or no data. These can still be great cars because experienced physics guys that I mentioned are good at filling in the blanks. You don't always NEED all the info to make a realistic car. Let me add that for AMS, even some of the cars we have 'officially' got in there, we got zero data. Sometimes I had something to help us from other projects but a lot of the time we had for example only a bit of telemetry. Sometimes nothing! Imagine the complicated aero on the Fiat Uno! Nightmare! :-)
@@NielsHeusinkveld I've...remade some cars in some simulators which I will leave unnamed for now ;), and it does baffle me when basic things are not making sense. When there is an indication of not understanding design, or mechanics for example. Things which are not excusable for the most part by lack of time/data. I find in this kind of situation, when it is largely down to the innate understanding of real car design and how the "numbers" usually are, where the ability shows. Tires are almost completely reliant on this quality IMO as there are so many unknowns that you need to "educated guess" very often. Consumers seem to have this kind of idea that the sim developers all get the perfect data and it only comes down to the accuracy of the simulation itself, but I would argue a lot is down to just the so called engineering design sense of the creator. Very difficult to make a fair estimation of simulation accuracy of a specific title when there is this factor affecting the result. With lack of data but a good "artist" you can make a better product than with all the data but an intern at the desk. ;) Of course it is just my opinion as a lowly modder.
Wow not even for Stock cars? Then probably you mostly didn’t have real drivers’ feedback too and based on racing footages. Yet many people say handling in AMS 1 makes sense, and some even praise it! But I think iRacing always uses real data, but perhapsnot super detailed - they might get simplified aero etc. Also I think AC devs said they were basing on real data at least to some extent. It’s also true for some official cars in rFactor 2 like Tatuus, and in RaceRoom.
@@JanKowalski-kb1hd It depends greatly Jan. For the Stock Cars in AMS we did get suspension data and telemetry which was extremely helpful. However they probably didn't even have aero data because of the low budget nature of the series. So we had to fill in some blanks still. This goes for a lot of small manufacturers who are still going by the seat of the pants feel rather than big budget simulation and aero testing. iRacing might not always have access to more data than other developers. It can be very hard to work with manufacturers and they won't suddenly hand over the secret data. Some manufacturers are extremely open, like I believe indeed Tatuus may well provide more data than others. I can't tell for sure, but the data a sim might have doesn't depend on the price or budget of these sims, more with the luck and communication skills of one or two guys on the team.
Enorm knap hoe je in een video van een kleine 18 minuten zo veel duidelijk kan maken en inzichtelijk maakt hoe gecompliceerd een tyremodel is. Erg veel van opgestoken!
Thank you! Your timing for this video is perfect, I'm well done with making tyres for AC after just over 12 months(I dont want to be a tyre creator anymore) you have filled in a lot of blanks, and confirmed some of my theroys. Cheers 🤓
Thanks for the video!! Always thought that tyres models are complicated and therefore oversimplified in racing sims, especially in console. PD: there was no watch in your wrist but anyway yeah, it was half-a-day video at least summarized in 17 minutes so well done!!
Niels that was awesome. Watched your last video and had these exact questions about the different types of modelling. Now I'm on my way to see if I can read (or watch) the philosophy of iRacing, rFactor2 etc. I am really hoping that with the influx of real drivers trying all these sims that we can have a push to make some positive changes! Also geeking out hard at your T2 in the background. I remember very clearly getting that and using my F1GP manual as a spacer to help connect it to my desk back in the 90s :))
Great video! I’m a chemical engineer but we did learn a bit about industrial process control in school. When you say you don’t have the “tools” to be able to tune a physical tire model, I remember that there were ways to perturb a model and develop its transfer function so that you would be able to isolate how to tweak the model to just impact one output without impacting other outputs. Do tire modelers use these Process Control concepts and simulation methods? It seems like what you need in order to be able to take feedback from testers and be able to then use it to adjust your physical tire model. Sorry for not being very technical in this post. It was a long time ago! I doubt I could do a LaPlace transform to save my life nowadays!
Reeeeeaaaally well explained 👏 Maybe a good point for the physical model is that if you succeed in modeling the tire from the ground, you can then give to your model the data for any type of tire and it will work out of the box with little effort. But yeah, easier said then done...
Yep, it is the holy grail.. Imagine if it worked flawlessly.. But that is almost certainly impossible.. I have more hope of using clever spreadsheets to fill in the parameters for empirical models and get good consistent results for any sort of tire / car.
The thing is each “sim” should be able to show for at least a nominal/variety of conditions that they have similar behaviour of the car they’re modelling in how it drives (would be great marketing material to actually show this, and I can’t believe it’s not done more). It’s got to a point where the community talks “tire model” and other issues so much I think it’s a distraction from what we actually want as the final output- a realistic, fun to drive simulator. In my opinion Dirt Rally 2 achieves this- no one, even the pro drivers, care which differential equation or model the designers used to achieve the output. I am an automotive engineer. Niels, Great video with great information for people interested 👍
Nice video. The problem with physical model is also that some tire types might be easily fitted by it but others might not. Real world tends to be more dampened than simulations of it and there are some calculation problems with that. You can get the basic right but then its hard to get to details like imperfections like uneven surface, compliance etc. These might have small effect but together changes the feeling.
Haha, that must be a Thrustmaster T2 in the background - I had a T1, which had rollers instead of pads for pedals (I assume to roll on your foot as you pressed down on it)... Gotta love that ancient tech - by today's standards, anyway, LOL. Anyway, I played plenty of IndyCar Racing and Nascar Racing with that wheel! EDIT: Yes, I just saw the USB conversion in your history... 🙂
Oh yeah, he's not kidding about when he says 'nearly everything'. :-) What a great book it is. Can read it a million times because I forget everything the instant I turn the page..
It's a fantastic book and his style of writing is also superb, if you haven't read "Notes from a Small Island" I'd highly recommend it, genuinely laugh out loud funny.
In the words of Private Jenette Vasquez (or therabouts) "Look man, i only need to know one thing......which is the most realistic feeling race simulation?"
A hell of a lot of parameters to calculate! Some devs seem to get closer than others. I’m very optimistic for the future of sim racing. It can only get better. Top vid Niels 👍🏻
Very interesting. thanks. My first thought is...I know you can model springs and dampers as electrical circuits. The sims need to do something the same so they can access to infinite tire tests that will help model the game 'tires'.
As a vehicle dynamics modeler myself I always take the Ockham’s Razor as a principle. Current state of physical tire modeling is too complicated with the lack of available data to make it work. Developers are exchanging accuracy for precision for PR reasons. That’s why still going with extended packejca in 99% of the time in sims will work better and more accurate.
Hey Niels, your video is awesome, so thank you for that. I would love to suggest you to invest 50 bucks on an average usb condenser microphone to improve substantially the quality of it tho! Greetings
Would one consider having more than one tyre model in the same simulator and then average them using diferente weights (or any other aggregator) ? Something like: calculateGrip( empiricalModel(carProperties, trackConditions, otherStuff), physicalModel(carProperties, trackConditions, otherStuff), someOtherModel(carProperties, trackConditions, otherStuff) ); calculateGrip function would apply different weights over the tyre model output. The weight for each model could be empirical, based on driver's feedback and car & track combination. I'm a developer but not on the simulator market. So this although might "possible" might not be feasible or correct for many different reasons. But, in my mind, having in the same sim a set of models, if viable, could help tests. In the non-simulator software development industry we can start a program and interact with it in an automated way, simulating what a user would do with their input methods. We do this to certify that after changing the system all the other features are still working as expected. Maybe, telemetry data could provide a way to compare models. Given the driver inputs, car and track properties and its current condition, could show where a given model "breaks reality" or how far from reality they are in different scenarios. I know, this sounds all "pretty", but by no means I'm saying it would be easy or even a good solution. I'm talking only as a fan and I recognise such "idea" can be half galaxy away from being useful or viable. All the best ;)
Niels is very knowledgeable and I have a great deal of respect for him and I also like the way he tries to explain things. Listening to him on this subject does bring up some points that I have been thinking about. That is, Niels is speaking from a position of being a modder. As far as I am aware, (and I could be wrong here) he has not been involved in the process of actually coding a tyre model from the ground up. Yes, he has taken A model and used it to create "tyres" to be used in a sim but he has not built a tyre model from code 0. So, albeit he is extremely knowledgeable he is also very limited in the understanding of actually building a real-time tyre model. Therefore, he cannot accurately assess how complicated it is for iRacing to change certain aspects of their model; he can only make an educated guess, (at best!) but without being privy to iRacing's code he is completely speaking in the grossest of generalisations. Yes, some cars in iRacing are bad at the moment, (GT3/GTE) but if he were to try the Skip Barber, Porsche 911 Cup car or even the HPD, these cars are actually pretty amazing at the moment. And, that's the thing, iRacing will keep chipping away and one day all the cars on iRacing will feel amazing, and most likely ACC, rF2, AMS2, etc... will all be long gone and replaced by something else that you have to purchase and re-learn.
Yes, good points, I'm making assumptions about these models and their ease or difficulty of use. I did hear from both rF2 directly and iRacing indirectly that tweaking is difficult, but I don't know the current state of the model or its ease (or difficulty) of use when it comes to iRacing. Then again, a physical model that you can tell "Do this to grip and load sensitivity when camber goes from x to y" wouldn't be a physical tire model anymore. So I do think by nature there will be less control over very specific tire force altering bits and pieces.
I've just stumbled upon your channel and your videos are awesome. I haven't raced online since the Hawaii days with Papyrus Nascar but I feel like getting a set of Ultimates for whatever reason. :)
Enjoying your videos ! Hey I have absolutely no business with rF2 ttool (maybe someday will, maybe never), but I can get some results while "working" with "realtime" section of TGM. Perhaps can't change that much in there, but for example I find that it is quite satisfying to control the way tire slips with those paramaters. I would say that it is even very intuitive. Many parameters for various kinds of static and sliding friction. Then bristle springs and bristle dampers are handy to adjust slip angles and sharpness of them in time. Temperatures stuff isn't too difficult to handle either. So there is lots of stuff that can't be "just tweaked" without killing yourself in there, but there also is quite some good stuff that makes it easy and intuitive to adjust IMO, and among that there is quite significant parameters.
Yes to some degree there is room to play with. I wish rF2 and iRacing really show us in detail how they go about as I clearly don't know the ins and outs of their models.
Oddly enough, those fun crazy experiments are actually a great test for a physics engine. And a great excuse to waste a morning. With RF1 for example I wouldn't have been able to make radio control 1:10 cars because of a bug. Its actually important in my opinion when working on a game engine to try silly extremes as they can point to bugs and weird things that normal cars don't always do. :-)
@@NielsHeusinkveld Yeah, I make small mods for my personal use from time to time, figuring out what does what is great fun! Although I got to say, sometimes those nasty limits are there for a reason - like unit limits in some rts games. Like star wars - empire at war. Turns out, merging all my fleets for final assault wasn't such a great idea, single digit fps kind of hurt. :D
The problem with a physical model is when you compare the output to real data, if something is off, the problem can be the model itself, can be in the data you are referencing, or can be in the data you inputed, that's trying to solve a Rubik cube that has only 1 color. you can hit a deadend doing everything you believe is right and then you have to change the entire model, without able to tell what was wrong with the current one before you start the new one....
I remember when the V7 tire model of iRacing came to the Porsche Cup and everyone was spinning like they have added rain to the game. But once I got used to it I enjoy the Cup Porsche even more than I did before. Also when they will add Rain this will make the tire model even more complex. Rumors are it will be this year when the official 24h Nürburgring race takes place, is this true?
It's just interesting how virtually every other sim got a decent tire model, yet iRacing feels like driving with wet tires on a dry track, not to mention rollovers that shouldn't exist. And all that from a company that is likely making more money than all others combined. If iRacing devs aren't lazy, they're really stupid.
Or, as an insider told me, Dave Cameron has the only word on what the tire model will feel like. Doesn't matter what the testers say or if they say it's shit (such as Billy Strange Racing or other real drivers), if Dave likes it then that's what everybody else will drive. If you ask me, I would go with the word of professional drivers, not a sim racing dad. GT racing? Ask Nicki or Dave Perel. Indy tire model? Ask real Indy drivers, there even a few champions that could help, and some trashed iRacing's tire model already.
@@amandasantini6265 "It's just interesting how virtually every other sim got a decent tire model" I see right about the same amount of people complaining about tyre models compared to sim X for every single sim out there.
If you're specifically referring to iracing, I get why people are complaining. iRacing is by far the most expensive sim yet has one of the worst tire models. People should be complaining.
@@Case_ For which sims exactly? I haven't heard nearly as much slander towards tyre models in any other sim. The latest I've heard was ACC but the tires models were updated, AC has good tyre models, RFactor 2 is miles above anyone in this category, Automobilista 2 is in early development, Project Cars 2 is the nearest in terms of how bad the tyre models are. So have I missed the communities feedback of these games or what?
The empirical model, while perhaps simpler and "less accurate", is perhaps more accurate as there is more room for error. If degredation can be input correctly, there is no reason why it shouldn't feel correct if each parameter can be adjusted based more on actual feel than hard data. For instance, in the movie "The Fly", Seth Brundle programs the exact data for a steak and his girlfriend comments that it tastes "artificial". If he'd have programmed empirical data, perhaps he could have built the steak based on his actual experience when eating one.
I’ve never heard it articulated before but I kind of intuited that there would be the two types of approach to tire modeling that you described, and to me the empirical model makes so much more sense. The physical model needs to deal with too much complexity and simply doesn’t have the data required to model the tire behavior or to judge whether your model is accurate. In the end, you’re really going for a subjective experiential result anyway, so why not make it easier on yourself? iRacing will never change their methodology anyway. They’ve got too much invested in their present approach and are too afraid of admitting that they were wrong to start over with a better process. The best thing they could do probably is hire Niels though!
Niels interesting video. Tire modeling is indeed a complicated issue due to the variances in real world tire data. I think the more reason for real world race drivers to properly validate the tire realism on sim the better. With that said, I am very happy on how most of the AMS cars handle on the edge of adhesion. Still a bit hit & miss for AMS2 but I hope they get to fine tune everything in the future. The wet weather modeling needs work, I think it's a bit too forgiving on trail braking in the wet and too grippy on the curbs, painted areas.
I'm having this Tyre problem in Assetto Corsa, if you look RaceSim Studio's F1 1990, if you try to get the qualifying time at Jerez1990 It's impossible as the tyres don't have enough grip. I did try other mod's tyres like from F1 ACFL 1994, they have enough grip to get the qualifying time, only by just. But I know I have too much grip, when I watch real life qualifying, I tend be going 10 to 15km faster. When I drop down RSS F1 1990 tyres, times around corners seem the same as real qualifying times, but overall still no way enough grip for the qualifying times. It seems to be me, Assetto Corsa needs more slip, less breaking grip, and high take off grip to solve this problem. It seems to me, that the way it is, if you have low breaking grip, you have low take off grip. If you have high take off grip, you have high breaking grip.
Thanks, I enjoyed it and didn't fall asleep once. Honest. Scienceguy some day: We now hold the understanding of all of the universe. Niels: Hold my tire.
I remember watching a video about racing sim physics and it was said that (at the time) all racing sim physics engines cease working when they approach zero due to some mathematical problem. Basically they said that when stationary or at very low speeds, you're actually using a different, "canned", model and that you then transition to the "real" physics model past a certain threshold. Is that true and is it true of sims today? I've not really researched the topic of physics in sims or computer games, but being born in the '80s I've seen physics in games become more complex over the years and got a sense of the empirical nature of it. I did wonder about physical modelling in a very naive way (basically visualising a tyre reacting to different situations) and it seemed quite intuitive, but the more I thought about it, the more I realised I have no idea how it would actually work from a mathematical point of view and what "physical" means in the context of simulation.
The best is F1 1991 HE. It is the only mod that has correct slip angles for F1 90's style, most sims model the Avon historic tyres relatively easy slip angles for historic racing. Actual goodyears had around 3-4 degrees slip angle
Where did you find this information and how reliable is it? Race Car Vehicle Dynamics has a slip curve for a Goodyear F1 tire where it maximizes around 5 degrees. Plus we don't know the test machine and conditions, there is a bit of a tolerance on such values.
@@NielsHeusinkveld 35 years of motorsports experience and 25 years of race sim experience. Yes there is tolerance due to heat, road conditions and scrubbing.
what about soft body physics for tyre simulation? Seems perfect, in my non existent game-engine experience. And friction forces, combined with that. The rubber is soft body phys. but the air inside is, well, air. Set the possible squishiness, reduce flex in or near walls... if the CPU power wasnt a factor, wouldnt that be the most relevant technique in theory?
These tyre models, rf2 and iracing, how hard is it for these game to compute in real time? Newgarden highlighted the tyre as the big difference between the real thing and iracing.How does inertia factor in? on track a tyre can lose all grip, lifting off the ground like in monaco or locking up or by slipping, but the car will always follow inertia, or the greater of the 2 forces, I wonder if emulating inertia and gravity is crucial to the tyre physics.
Great video. Just wanted to add this ressource, where ACC's Physics developer Aris tells some stuff about the tyre model in ACC: www.assettocorsa.net/forum/index.php?threads/introducing-the-5-point-tyre-model-for-acc.59307/
I'm a mediocre sim racer, but a decent real one in my times, the most realistic tires model for me is in the live for speed sim (yes that). Y tried every sim posible.., but lfs is the most "this is like for real feel". In the other sim i have this sensation of "it's more difficult than in reality" i m not used to spinning around like a fool at 50mph...im not says that is the more accurate possible, but it simulates a feeling of realism
I've been involved in tire modeling and multibody simulation for the formula sae team of my university... I can say that tire modeling is complex, but is not a magical thing. Tire manufacturer knows how to measure tire data but they don't spread this to all the world;They can publish the vertical stiffness and dumping but usually they stop here. For the modeling then there are a lot of ways to do it, the most used is the "magic formula", by pacejka, that I think is a good compromise between realism and complexity but i don't think that it's used in Sim racing. However I know ( from an interview of Giampaolo Dallara) that pirelli can model a tire that is capable of replicate even the tire degradation. For a real professional simulator, like the dallara Sim or f1 sim, the tire are the biggest problem: more than a alf of the parameter of a model are for the tire. I know that I'm talking about another tipe of Sim, but whith these I just want to say that nowadays, whit the right hardware, we can symulate something that is really really close to reality, but I think that it require to much resources compared to what can give in terms of enjoyment.
Hey Niels, I appreciate your remarks about the practicality of 'empirical', curve-based tyre models (ala Pacejka) over 'first-principle' models (such as in rF2, iRacing, PC2/AMS2). However one question I have is how do you deal with the finite size of the contact patch? ACC initially used a 'single-point' contact patch which led to some issues on for example curbstones, and transitions between tarmac and grass. Now they switched to the 5-point contact patch and that already seems to make a significant difference. Do games like AMS1 still use a single-point ('infinitely small') contact patch? Is there any way to include the 'averaging' effect of a finite-sized contact patch without resorting to a finite element/finite difference model like in rF2/AMS2? Alvast bedankt ;) Cheers, Gillis
Fascinating video there, thanks for taking the time to put it together. You have me wondering now whether AMS2 is using more of a physical model rather than the empirical one you are more experienced with, and if that might be one of the reasons for you moving away from AMS2 development.
Yes indeed, the tire model would be too big a challenge to learn, and it would be too uncertain what the outcome would be. If I was a student again (i.e. more time to spend ummm 'studying').. I would like to devote a few months to it or to rF2. The physical models ARE fascinating, and I bet I'd learn more than a few things studying one. I doubt it will be my choice for a racing sim. Nowadays aged 40, time flies and there isn't enough of it. It feels like a big risk investing a lot of time into it and I fear it won't be as worthwhile as putting the time into other things.
@@NielsHeusinkveld I can totally relate to that. When you have spent a large part of your life building up knowledge and experience on a subject, you can look back and appreciate just how big a mountain that is to climb. It's not to say you couldn't do it again, and do it well, but as you go through life you have to pick your quests, and it has to be something that captures your energy and imagination. I appreciate such an honest reply in such a public place, and I'm sure whoever gets the benefit of your energy and expertise will be all the better for it. Onwards and upwards old boy :)
With the current situation, so many of these IRL driver races. And quite a few of them criticising these tire models. Can pressure now be applied to the tire manufacturers to measure more thoroughly? At least giving the physics guys and sim studios more numbers.
Tire manufacturers don't give a skidmark about sim developers I would say. Until a sim developer shows up with a big bag of money to pay for the testing..
3 minutes in and thoroughly educated already. Just one suggestion, is it possible to reduce the echo somehow? Perhaps wearing a headset (I understand that might look "weird") or putting up foam on the walls would help
