Assetto Corsa Competizione: The 5 Point Tyre Model Blog

Paul Jeffrey

Premium
ACC c.jpg

In another very interesting post from Kunos Simulazioni, physics developer @Aristotelis talks about the upcoming new 5 point tyre model for the simulation.


Developed by the makers of the popular Assetto Corsa racing simulation, and benefitting from the official licence to represent the 2018 and 2019 seasons of the Blancpain GT Series, Assetto Corsa Competizione continues to be expanded and improved upon following V1 release earlier in the year - the next development set to be a new 5-point tyre model...

Five point tyre model I hear you say... pray tell, what might this be? Frankly I've no idea, so I'll leave it to Aris to explain...

In modern computing, collision detection is still one of the most resource consuming operations a physics engine has to do. It needs to be very fast, very accurate and low resource demanding, but you can only choose two of them at once… Add to the equation that in AC and ACC the cars drive on an invisible physics 3D mesh with millions of polygons, derived from the laser scan of the real circuit, and it’s clear that a compromise has to be made. On top of that, AC and ACC uses the same player physics for the AI, requiring even more resources.

The collision detection of a simulation has to be very fast and very accurate, otherwise strange things might happen to the simulation. So in order to make it less resource demanding the tyre model of AC and ACC uses one single point to determine contact of the tyre with the terrain, being this last one a flat asphalt, bumps on the asphalt, various types of curbs, grass and so on.
This solution is a quite good compromise in order to have decent performance and high simulation accuracy. It permitted us to push hard and evolve the tyre model, adding more and more features on it. As a matter of fact the ACC tyre model is one of the most evolved ones, completely dynamic with a wide range of causes that affect and influence the grip and response of the tyre. Various heating levels, different tyre wear features, various influences in tyre rigidity and damping, completely dynamic slip ratio and angles, dynamic rolling resistance in different situations, full water draining simulation etc. etc. the list is very long, very complex, innovative and often involving completely original and breakthrough solutions, derived from meticulous studying and hard work of Stefano Casillo that you won’t find in any scientific paper, as he had to build new equations by himself.

Unfortunately, while still acceptable in AC, the more advanced physics engine of ACC put in evidence the limitations of the above solution. Our tyre model started to have issues and downright buggy behaviour under certain conditions over curbs. The use of laser scan circuits, gives no doubts on specific features of the circuits. If a curb is high, has a specific angle, has dangerous angled steps in it or any other strange feature, then the laser scan will show it in millimetric accuracy. On the other hand, our company motto is that we take no shortcuts in things we are certain. If a curb is made in a specific way, that’s how that curb is going to be implemented in the sim. If that means that our tyre model is going to suffer on it, then so be it, we will have to work hard to make it better and for sure we ain’t going to make the curb smoother just to “workaround” the issue. So let’s analyse what exactly happened with our tyre model in such conditions.

First of all, let’s see how the tyre model would deal with a high but smooth curb, like the many of the curbs at the circuit of Spa Francorchamps. As you can see in the screenshot a curb like this, although it has a smooth surface, it has a quite steep angle, often exceeding 30°.

ACC a.jpg

The single contact point is approaching the curb but still stays in the completely horizontal surface of the asphalt. Even if graphically the tyre starts to be “on the curb”, the actual point still remains down on the asphalt. This also tricks the mind of the driver, because in real life, if the edge of the tyre touches the curb surface, the driver will hear and feel the tyre touching the edge and take appropriate action or at least he will know he is gradually going over the curb. In ACC this won’t happen. As an example, many people see in real life the left inner curb of Eau Rouge being dirty from tyres and think real drivers abuse the curb. They try to do so in ACC and get an instant spin. In reality, you just need to touch the curb a bit and you can feel it and make it dirty. In ACC the driver will probably think he still has space and on the next lap he will go even close. This creates also a false impression to the driver of where the limits of the car are and the driver might caught himself trying to place his wheels in places he shouldn’t do. At some point the single point will go over the curb. Instantly it will understand a 30° surface inclination. This is an extremely big change on the contact point and a huge spike in load, vector, rolling resistance and so on, resulting also in big spikes in forces and grip. Consider the following situation:

You are on the limit of grip in a turn. The outside wheels are right on the edge. You climb with your front internal tyre on a high curb. This means that you raise the front inner end of the car and obviously you load the rear tyre with even more load. The tyre, already at the limit of its adhesion, cannot afford any more load so it starts to slide. In AC/ACC the front inner tyre will also take a big spike in load and rolling resistance, so it actually brakes for a moment and throws to the suspension more forces than it should. Those forces end up to the rear suspension and tyre in big spikes and the tyre loses even more grip. Usually in very stiff racecars, the inner rear tyre might even go airborne losing all grip and forcing the differential (if locked) to move even more torque to the outside rear tyre.

ACC b.jpg

If during the whole process, you also remain on the accelerator, you will have a situation where the rear outside tyre pushes forward with less lateral grip while the front inner tyre pushes backwards. Practically you car is transformed to a tank with treads that move in different speeds.

ACC c.jpg

This is why in AC/ACC if you modulate your accelerator the issue is practically non existent but also why if you stay on the accelerator the behaviour is exaggerated.

Another example are stepped curbs. As you can see in the following screenshot of a Paul Ricard curb, the steps of the curb are not equal from side to side, but gradually become more and more deeper (or in other curbs they might go upwards).

ACC d.jpg

The tyres of a GT3 car are quite wide, around 30cm wide. This means that often the tyre can be over the curb with the outside part going over the more shallow part of the step, and the inside part going over the completely flat part of the curb, leaving only the center of the tyre hanging through the most deep part of the stepped curb. The result in real life is that while you will feel vibrations from the curb, they will not be as important as the most profound part of the curb. Again, unfortunately in AC/ACC the single contact point at the center of the tyre, will get the worse possible condition of the stepped curb, something that in reality would never occur. So again in AC/ACC the behaviour is much more harsh and critical than in real life.

Finally, some kerbs have an almost vertical step at their outside edge.

ACC e.jpg

Often the driver will ride and go over the edge of said kerb and then slowly return to the main road going almost parallel with the curb. ACC in particular is extremely critical in such situations. Adding full 3D flex of the contact point (only vertical in AC), created a critical condition in the above scenario. The single contact point would go to the vertical parallel side step and being as high and vertical, instead of climbing over it, it would start to flex outwards, practically getting trapped in a rail. The driver would see that the car wouldn’t follow his commands to reenter and at some point he would move some more the steering wheel, creating more lateral force than actually needed. The front tyre contact point would climb over the step and then obviously would have excessive slipangle that would steer the front end very fast. At the same time, the rear tyre would be in the same condition and still trapped, so it won’t be able to follow the rotation of the car and will continue straight ahead in the rail, practically inducing the car into a very fast spin.

ACC f.jpg

This is one of the most well known and widely reported “curb of death” situations in ACC and in great need of a solution.

Paradoxically, sometimes trying to make the tyre model even more accurate, detailed and realistic to drive, you get into extreme situations that are so critical and have so bad side results that can practically destroy all the good intentions and effort to offer an even better simulation.
Fernando has been hard at work on our tyre model. With the help of Stefano so that he can understand the underlying code and make sure we get no big impact in performance. Some extra code performance optimisation from Fabio was more than welcome too. So from version 1.0.7, ACC now features a 5 point contact model! We implement 2 contact points at the edge of the front of the tyre footprint, 1 in the middle of the footprint and 2 more contact points at the edge of the rear of the footprint. Each single point moves and flexes independently reacting on forces and surface contact, but also, predictably forces to move the other points together, averaging the resulting forces and vectors, giving a much better representation of what an actual tyre would do.
Examining the above 3 examples again, we can observe massive improvements of how the new tyre model is reacting.

On the first smooth high curb situation, the advantages are multiple. First of all, when the edge of the tyre touches the curb it activates the sound and properly moves the FFB steering wheel, thus communicating at the driver the correct width and position of the tyre. Furthermore the contact points at the edge of the tyre, get the spike of the steep angle of the curb, but their forces are averaged to the rest of the contact points that are still on a flat surface. The tyre actually “climbs” over the curb, instead of instantly finding itself on top of it. There are no more load and angle spikes except the realistic load changes.

ACC g.jpg

So obviously if you are too aggressive the rear tyre will lose grip and can still provoke a spin, but the result is no more exaggerated so there are no more excuses for bad driving ;)

On the second example, the contact points now include the whole width of the tyre and if that’s the case successfully keeping the middle of the tyre in the air while also being spread longitudinally in the footprint length. There is always a contact point touching the surface at the front or the rear of the footprint even if the tyre is rolling on the steps. On top of that, the extra points are controlling for load spikes and avoid situations of excessive rolling resistance or vectors that point backwards to the car motion.

ACC h.jpg

This greatly improves acceleration over stepped curbs, as in example at the exit of turns, which in the past, drivers would avoid in order to not harm their acceleration.

Finally, on the most important third example, as clearly explained before, the multiple points now permit the tyre to “climb” over obstacles. So when the edge of the tyre hits the vertical step of the edge of the curb, those contact points start to flex and go parallel “entering the rail”, but the rest of the contact points, still push through the direction and push also the edge points to climb the edge. The driver doesn’t have to do anything with the steering wheel, and the tyre simply goes over the edge of the curb without any dramatic situations.

As an extra bonus, we have also added a new dynamic feature to the tyre flex behaviour. As you know the footprint of our tyre model now flexes in three dimensions. Going even further in tyre simulation, the lateral flex provokes the tyre to lower its profile. This means that the more the tyre flexes laterally the more the ride height lowers. Obviously the change in ride height is minimal, but in a car with proper simulated aerodynamics we know that even one millimetre is important to the handling and balance and so this new tyre model feature plays an important role to the car’s handling. You might notice a bit less understeer on power exit with some rear and mid engined cars, since the lateral flex of the tyre will bring the nose very slightly lower. Gives a bit more control to the front end of the car.

The new tyre model feels even more accurate to drive, permitting placing the car with more precision and absorbing bumps and undulations better with more predictability. All of this is now possible without any hits in terms of performance, which is practically a miracle and win win situation for all of us. As I’m writing this article, we are working very hard on balancing the various values and testing performances and handling, so that laptimes will remain more or less equal and general balance of the cars and your setups won’t change, except maybe for better precision, stability and predictability of the tyres both on and off the curbs.

One more thing…
In our never-ending research for more realism, we knew that we had to improve our Traction Control systems. With the strict rules of the Blancpain series and the BoP in place, there’s not much performance to be found in terms of aerodynamics, chassis, engine and suspension. But the electronics, although regulated, are still a big open field and automakers spend a big amount of their budget to improve such systems.
In street cars, the Traction control engages on the aperture of the accelerator (drive by wire), on the engine ignition timing, on the engine ignition cut and on the brakes. With a combination of all of those controls, the engage of the TC can be smooth and highly efficient.
In GT3 race cars, the control of the throttle and the brakes activation is prohibited by the rules. The TC can “only” modify the ignition timing which lowers up to a point the engine power delivery and if this is not enough, then it will start to totally cut the ignition at a high frequency, resulting in the characteristic engine rattling and vibration.
Been able to lower the ignition timing before cutting totally the ignition, is very important because it permits a more gradual power delivery and a more accurate control at the initiation of a sliding, giving the driver the possibility to modulate throttle application or work around the slide with steering inputs.
To better control the TC engagement, the ECU not only tries to estimate the sliding and slipping of the tyres, but now also uses gyroscopes that measure the yaw rotation of the car, sensors on the steering wheel, and many other parameters so that it can accurately calculate the acceleration of the yaw rotation of the car and understand and predict if it is controllable by driver or if it has to intervene to slower such rotation and give the driver the time to react.
All of the above is now simulated in ACC, delivering a more advanced TC intervention that not only can make you faster and safer but also, incredibly enough, make the car handling more enjoyable even over the limit! Honestly I never thought I would say this for a electronic system…

All of the above, was not a small feat by any means. It took hard problem solving, analysis, non stop testing and balancing. We are very proud of the results as we think they push the simulation realism even further and we are confident that all of you will enjoy ACC even more and appreciate the efforts to not deliver you just another racing game, but really push the boundaries of simulation. As with such complex systems, we hope everything will work flawlessly but if you find any issues, please report everything to our support forums and we will do our best to resolve them.

Original post HERE.

Assetto Corsa Competizione is available now on PC.

For more from the world of ACC, why not head over to our Assetto Corsa Competizione sub forum and get yourself into the thick of the action? We have a great and knowledgeable community, plus some pretty epic League and Club Racing events, if I do say so myself. Go on, treat yourself!



Like what we do at RaceDepartment? Follow us on Social Media!


 
 
Last edited by a moderator:
Wait, so now you admit that you are a fanboy ... and then ask where the bias is coming from? Then try to turn it into "deserved support". No further questions your honor.
This is getting a little personal for my taste and I don't think we are understanding each other very well, I like game and disagree with your opinion, give constructive feedback not generalised complaints that are no good to Kunos, your experience is not the same as other's.
 
  • Deleted member 379375

I'm looking forward to testing differnt cars again with a view to possibly changing my go to car, currently the Ferrari. Tried the audi on spa for a while and it was rediculas to be honest. One question I have is whether the change has a noticable effect on car behavior when not on curbs?
 
Why do people who dislike the game feel the need to come onto a topic about the game just to say that they don't like it? Then as soon as someone disagrees with you they are a 'fanboy'.
Its not just here either. I have been on other forums and as soon as an iPhone is mentioned you get someone coming on saying how rubbish they are and how much better their Android is. You mention the Rift and someone will come on to say their Vive is better.
Please people, let us who enjoy our games/sims 'enjoy them'. We don't need you to come on asking us to look at reviews for a game that we have already spent many hours playing and enjoyed (and still enjoying) every minute of it.
Its as if you feel threatened for some strange reason that this is going to be the best sim out there, and you have to hunt down threads to post your hatred.

As for the new tyre model, this looks great. Ok, it may not be new, but it is new for ACC and I for one am looking forward to playing my favourite racing title with the new updates.
 
Last edited:
@kunos / @Aristotelis

Cheers for the excellent writeup! :thumbsup:

Am I to understand that the point (pun intended) here is that there are now 5 brush elements in contact with the ground at all times (or, well, most of the time), simulating the edges and the middle of the contact patch in an (elongated?) quincunx pattern?

Something like this when viewed from above (or below) where each point represents the end of a bristle on the brush, which is connected to ... the rim (for purposes of calculating height under 3d deflection)?

Code:
front (direction of travel)
^^^^^^^
o     o |
   o    | contact patch length
o     o |
-------
contact patch width

Hello and thanks!
Yes the tyre model works more or less as you described.

With PC2 (and I guess PC Pro and AMS2), I believe SMS settled on a grid of 3x11 points per contact patch with a physics calculation rate of 600 steps per second, which obviously has the drawback that the engine simply can't run the same model on the AI cars (I saw some tests being done with a CPU core per wheel, which helped mitigate the performance cost at least a little). I also saw tests with more points per contact patch and more simulation steps per second, but they weren't workable on console CPUs AIUI.
Obviously I cannot comment on other sims as we don't access to the source engine of course. Many devs do many amazing things out there for the genre.

Your brush model also appears to be quite advanced (aquaplaning etc.); do you also simulate heat propagation from surface through the various elements of the tyre/rim/air/brake setup in any way, shape or form?
One of the main reasons we stick with our very customised brush model, is because it is relatively easy to work with and gives results very fast. Practically change a value and drive. In an research era (tyre simulation) that still there are tons of unknown data, difficult to get info and a lot of reverse engineering, for us this is a big advantage. I'm not saying that our solution is the best out there, but for our workflow, it suits us very well. It also permits to add many completely original and customised code that the original brush model doesn't include. You have already mentioned our wet simulation. This is based on a film of water that depending it's depth the tyre can or cannot drain. So it's not just less grip, you actually have plenty of grip considering that rain tyres are even softer than normal slicks), but the capability of the tyre to drain the water depends on the depth of the film, the speed, the load, the design of tread etc.
Other customisations include graining, blistering, flatspots that AFAIK we were the first to include in nKpro (flatspots) and later AC. The tyre wear affects the radius (separately for 3 circumference zones, so camber will affect different zones of the tyre so to say) as well as heating and slipangles/ratios of the tyre and flex, stiffness and damping of the footprint. 3 levels of heat management of the tyre, 3 circumference zones for surface flash temperature, core heat, air heat. The latest is also influenced by the brakes heat and of course each one influence the other. Again all heat influences not just the generic grip and pressure of the tyre but also stifness, damping, slipangles/ratios dynamically. The flash temperature is also affected by the wetness of the road and so on and so on...

In terms of AI cars, do they also run full player suspension/aero models in AC and ACC or do you have to simplify them a little due to physics calculation budgets?
Yes they do run on the same suspension, drivetrain, engine, same aero models, everything. Not sure if at the core Stefano might use fixed aero setup to avoid AI instability at times. Generally I make them use the same safe setup preset of the player and they handle it. In terms of calculations, they use the same everything, but still on single contact point.

Finally, how many times a second do you do physics mesh collision calculations for the contact patch?
333hz

Inquiring minds would like to know. :)

All the best with ACC!

@kunos who is of course the mastermind behind the whole physics engine, can confirm my writings or put me in shame and give far more accurate information :D
 
@kunos who is of course the mastermind behind the whole physics engine, can confirm my writings or put me in shame and give far more accurate information :D

ya pretty much correct, aero is the same and they seem to be able to cope with it fine.
On the top of my head the 2 things AI physics does differently are:

- No tyre flex and soon no multipoint collisions for performance reasons
- No "drivetrain flex" because of performance reasons and because it messes up their formulas for throttle application
 
Is the AI running on the same tickrate, and if so is there any advantages to it / would you consider to lower the TR for better performance ?

yes they do run at the same tickrate.
Having them on something different would require a major redesign of the entire physics subsystem with results that are pretty difficult to predict (ie. the actual gain might not justify the complications in the code)... so it's unlikely to be something we'll go for with ACC.
 
yes they do run at the same tickrate.
Having them on something different would require a major redesign of the entire physics subsystem with results that are pretty difficult to predict (ie. the actual gain might not justify the complications in the code)... so it's unlikely to be something we'll go for with ACC.
So no chance of epic 70 car grid at Spa? :(
 
  • Deleted member 113561

This is getting a little personal for my taste and I don't think we are understanding each other very well, I like game and disagree with your opinion, give constructive feedback not generalised complaints that are no good to Kunos, your experience is not the same as other's.
Finally, a comment from you that makes sense. I gave constructive feedback on the Kunos forums but also made them aware that I think it is not okay to release a game in an unfinished & buggy state. Some steam reviews have a whole list of specific complaints, I think at least half the people had the same experience as me, many just react differently and move on, instead of complaining.
 
Last edited by a moderator:
@kunos / @Aristotelis

Thanks for the detailed comments to both of you -- much appreciated! :thumbsup:

With the advent of more and more advanced physics engines, it seems to me that modding and tinkering is becoming more and more convoluted.

Do you think it would be feasible/possible to do a backport of (possibly only parts of) the ACC tyre model to AC (to the benefit of modders), while still keeping the newer features (graphics engine, weather, official licence etc.) as Unique Selling Points for ACC?

It'd be pretty awesome if AC could live on as a sort "long tail", grass-roots modding engine and essentially replace rF1 on the modding scene. The 5 point contact model sounds like it'd be an ideal fit for AC too.

P.S. You mentioned the contact patch collisions are calculated at a frequency of 333 Hz. For your particular use-case, what benefits/drawbacks are there to going higher (if any)?
 
Last edited:
Obviously a great improvement that will help ACC exceed AC, even if some other sims already have something similar to this for years.

Not sure if I am missing something, though, but when does it arrive? Or is it already in the most recent update?
 
I'll answer these two questions by @ermo and then stop.. let's not turn the thread into a Q&A :p

As already explained way back in our forum, the situation on the AC front is pretty complicated right now. An update from us will most likely break all the various "aggressive mods" that have been created in these years.
This will create an uproar of complaints and conspiracy theories that I am happy to avoid by simply not doing it.. I've been around this community long enough to know how it operates and functions... I still remember us closing the modding section on our forum because it was taking too much time to check the actual legitimacy of the content presented... and then see what if felt like an unstoppable flow of conspiracy theories threads appearing there, here, on reddit and everywhere else. Been there, done that.. not going to do it again.
Actions have consequences, it's true for us but it's also true for the community.

Having said that, the NUMBER ONE reason why this is not happening is because we're all focused on ACC and all hands are at work to make it the best product we can... the point in the paragraph above is 100% personal, we never even sat down to consider back porting physics to AC.. I personally back ported a couple of stuff for fun over weekends... but with no real plan for release. Back porting other things would mean to involve developers that were not even in Kunos during the AC days.. completely unfamiliar with AC's codebase and, because it won't happen during own times on weekends it'll have to be a coordinated company decision, not just one guy hacking by himself.

As for the frequency.. it's a balance between accuracy (the higher the frequency the more accurate the numerical simulation is) and performance (more cycles = more CPU needed). Of course this should never be used as a meter for actual overall simulation accuracy among different engines... frequency is just a number, what it really matters is what's done in those 333 updates.. so the above "the bigger the better" is only valid in the example of the same engine running.
It's also interesting that, when dealing with 32bit floating point based engines, going too high on the frequency might actually start to create problems because of the 32bit resolution of the numbers... in ACC this wouldn't be a problem (on PC) because we're using 64bit double precision for the physics.. but still and interesting fact and a good way to plug one more difference between ACC and AC... :p
 
Back
Top