Jump to content Jump to content
F1 2021 | Online Services Down For Maintenance | 26/07/2021 ×

Let's Talk About Tarmac Physics In-Depth


Recommended Posts

TL;DR

  • No, tarmac physics is not just gravel physics with more grip. It's more complicated than that.
  • Yes, there is always room to make tarmac and car physics more realistic. Any sim game does.
  • If you have trouble with tarmac handling, check the car setup if it is the same as monte carlo and Germany. Adjust brake bias and pressure so you aren't locking up the front wheels. Reduce diff lock strength. As a last resort or for testing purposes, turn on ABS and traction control to understand where the car's limits are.

 

“A rally stage is nothing but a series of near crashes.” - Damian Harty, Prodrive Engineer

Is there any more controversial topic in sim racing than tarmac physics? Whether it is dirt rally, ACC, or iracing, any time a sim racer decides to update their tire model, a flood of complaints come rushing in, drivers foaming at the mouth at the lack of realism and unpredictability/predictability of what’s happening underneath the car. But these models are such an important part of a sim racer’s experience, and along with all the visuals and sounds, play a big part in the driver’s immersion.

In this exploration of the topic, I hope to illuminate how a physics model is put together, the challenges involved, and then take a look at how Dirt Rally handles these challenges.

First thing we need to look at is how a simulator models car physics. You need to model three main components in order to accurately portray vehicle dynamics - a vehicle model, tire model, and a track model. For Dirt Rally and most commercial racing games, they would be described as a dynamic simulation, meaning that “the effects of inertia and
shock absorber configurations” plays a role in the driving experience (Segers, 402). You simply cannot model a gravel stage without detailing what’s happening in the dampers and with weight transfer. Take the fact that a rally car on gravel “only has its wheels on the ground 50% of the time,” as stated by former prodrive engineer Damian Harty (Abid interview). If this is not portrayed in-game, you don’t really have a great model for gravel.

Each of these components have the option of varying complexities, and the developer needs to be able to balance the computational power and storage space available to process the physics in real-time (Beckman). As such, racing games will vary in their ability to provide complexity, along with balancing the game’s playability and immersion.

Let’s look at the first component, a vehicle model. You need a vehicle to fully detail how forces on the car are transferred to the tire. A full vehicle model would detail the height of the center of gravity (CoG), its relation to the roll center on a car (the point where the suspension body ‘rolls’ the car around), the weight of the sprung mass, the weight on the unsprung mass, and so on, and so on. You could find a litany of things to simulate, but each one of those components increases complexity by any number of degrees. Take the example of beamng, or its predecessor, rig of rods - this soft-body physics model takes into account this rolling effect, and even how the suspension bending at the front is transferred to the rear of the car. Very impressive! But it comes at a massive cost - the game has very heavy computing usage and if we are being honest, also comes at the cost of its visuals. 

So what can be done instead? Enter the simple point mass vehicle model (Segers 402). Everything about the car is reduced to its single point, the CoG, and the height assumed to be at the road level. You then add whatever forces you want acting on the car to generate on this single point, so you aren’t dealing with multiple points and thus exponentially increasing computations. You can simplify a lot of the forces acting on the car by doing this, it’s computationally cheap, but it has inherent weaknesses that you have to further simulate the effects of body roll and it can lead to weird things happening in the game engine. One theory is, this simple point model explains why a lot of players experience this weird steering effect in Dirt Rally and other arcade racers where it feels like you are steering from the ‘center’ of the car, or other weird behaviors as described by forum user BayouRally). 

This leads to one of my first theories about the nature of DR2.0 physics - aerodynamic downforce is not modeled. You can see in this graph (https://i.imgur.com/lC6Zokd.png ) lateral G forces plotted against speed for the ford fiesta mk2 in spain. If aerodynamics were modeled, you would see this graph flare out like a trumpet as speeds increased as downforce acted on the front of the car, adding grip to the front tires. But this does not seem to be the case as evidenced by the data. Rarely is this mentioned in discussions about the game’s physics, and I can only assume that as far as immersion goes, not many people have experience with how aerodynamics works or feels in real life. ACC and iRacing do model aero, and to not have this aspect of the game modeled leaves out a very big chunk of realism.

Moving on now to the tire model aspect of sim racing. You need a good tire model in order to form the basis for skill-based driving. The tire model takes the forces produced by the vehicle model and turns them into the forces necessary to get the car moving, thus forming the friction circle that will encompass the limits of your car.Tire modeling, unfortunately, can be insanely complex. The Pacejka Magic Formula tire model comes up often, but is useless for any real-time simulation.

Solutions do exist outside of this particular model, Brian Beckman made a simplified version of the formula that I believe has been used by game developers since the early 2000s, and then there are lap time simulation designers like Danny Nowlan who have come up with even simpler methods that let you design your own tire model without the need of tire manufacturer data, and can use your own data or hand calculations and so on. But it is important to note that there is no consensus on what the best one is. So choosing a model should be made with respect to how all these models behave and is appropriate for the simulation you are designing.

This is all to say that tire models are complex but necessary, and it is difficult to ascertain what model Dirt Rally chooses to model its tires. You also have to remember that off-roading is often operating at a 3D level, so the tires work differently depending on how much the surface is changing underneath it. Plus, each car is different and develops their own unique friction circle. Very tricky stuff! But luckily, we can look at some data to see roughly what is going on with DR’s tire physics. We can observe one form of longitudinal slip angle on tarmac and gravel by examining how the car operates from a standing start. Here’s the Porsche RGT burning some rubber from a standing start on Spanish tarmac and New England gravel. We are looking at longitudinal slip ratio, which is just the average speed of the driven wheels divided by the average speed of the free-rolling wheels in a RWD setup. When the slip ratio is positive, the driven wheels are faster than the free-rolling wheels, and vice versa when the ratio goes negative. The yellow line is gravel while the white line is tarmac, measured by distance along the x-axis. No surprise, tarmac grips up fast, and gravel lags, as you would expect it to. The tarmac has the quality of being slippy until it’s not, with a very fast drop to its ideal slip ratio of around 15-20%. The gravel takes a while longer and gradually forms its ideal slip ratio in a higher range, 30% or so. You can think of it like going in the opposite direction of a chart like this one, even though this is only ONE aspect to what the tire is doing, so do not take the graph’s numbers, or the logged data for that matter, as gospel. Remember that you need separate tire models for tarmac and gravel as both are completely different kinds of tires.

This isn’t all that surprising when you think about it, and it conforms to what you would expect driving in an off-road setting versus tarmac. What is interesting, however, is the amount of ‘chatter’ that is occurring while the slip ratio converges to its ideal, but we’ll come back to that when we talk about track models. 

For lateral slip angle, this is trickier to find. I have demonstrated in a previous post how we can take the car’s yaw rate and angular velocity to get its attitude velocity, which is somewhat of an approximation of actual slip angles during lateral movements and goes positive during oversteer and negative during understeer. When we average the logged values for attitude velocity, we get a number that represents the car’s tendency to mechanical oversteer/understeer. We can also look at how oversteer/understeer affects the accumulation of G forces. If we add both longitudinal and lateral G’s together, we can track where and when we lose grip in particular driving maneuvers. If we look at how this works on tarmac, for instance, as the car develops more oversteer, there is a correlated drop in the amount of grip available. We are, in a sense, past our peak grip levels and losing performance, potentially losing control of the car. We compensate with driving inputs to bring it back to normal. This effect changes somewhat when we look at how this works on gravel. First off, we simply do not generate as many G forces as we do on tarmac. But, we can see that when we develop oversteer, we do not see the same correlated drop in grip - this suggests that the gravel physics are giving us more lee-way with our peak grip, or that we are developing a certain amount of lateral grip as the car drifts. The amount of slip angle we can develop is much more shallow than on tarmac (examine the shallow drop into understeer as we accelerate out of the hairpin - no significant loss in grip during the whole event). We might not be taking advantage of the full grip available, but rally driving spends a lot of time outside of the friction circle to position the car for better exits and racing lines. 

This is more easily seen when we plot attitude velocity against lateral Gs. This graph shows the ford fiesta mk2 on tarmac, whereas the operating band for developing maximum lateral Gs is narrow, a limit somewhere in the range of +/- 30 deg/s. On gravel, we are given a wider band to develop maximum Gs in the oversteer direction, despite the lower grip, about +50 deg/s to -20 deg/s. Not to mention that with the reduced grip, we are operating in a space with much more oversteer and understeer than on tarmac. Now, for tarmac, if we were very good drivers (which I’m not) we could maybe try and find if there are maximum grip levels to be found somewhere between 0 and 30 deg/s, but frankly this requires a level of skill that I simply do not have and the fidelity of which I lack with a controller. This is why in the car setup, we try to compensate for these oversteer/understeer problems by moving brake bias to the rear, reducing brake pressure, and fiddling with the diffs so that we are not creating large slip ratios and slip angles that lead to control issues. Doing these setup changes is the kosher way of managing slip issues without resorting to ABS, traction control, and stability control assists, even if those assists do find themselves in modern motorsport.  

For our last component, the track model is what your vehicle and tire models are driving on. Without the track model you cannot have a dynamic simulation that is constantly changing the way the car and the tires behave beyond their stated design - what fun can be had driving a fully detailed vehicle and tire model if you are driving on a surface that never changes? Aris Vasilikos, who works on Assetto Corsa’s physics, talks about how big a difference laser-scanned tracks meant to the sim racing world (Abid Interview). Being able to scan down to the millimeter every little dimple and bump on a track gives the car’s suspension system an endless amount of variations that the driver has to contend with, and it is what gives each track its character.

As too with rally, if not even more so, as we are not simply dealing with crests and bumps, but also dynamic changes in friction on loose surfaces. These variations are why rally dampers, for instance, are set up the way they are (I will be making a post about dampers in the near future that will go into detail about this). It is also where car setup really starts to make a difference in how the car model interacts with the track model. The ‘chatter’ in the slip ratio traces and as oversteer/understeer develops, for instance - this plays a fairly large role in how the surface interacts with the car and tires. What you are seeing are the tires speeding up and slowing down as they find traction on the gravel. This does happen on tarmac too, but on a much smaller magnitude. This constant back and forth is modeling what tires on gravel surfaces do, which is digging into the dirt and trying to find grippier surfaces underneath.

Now, we have to consider that Dirt Rally’s surfaces are not laser-scanned but instead crafted by hand, and that the surface model is quite consistent throughout the stage, with minor variations coming from the level of gravel (light, medium, hard gravel) or the type of tarmac (cobblestone, dry tarmac, ice). The micro-variations on the road surface are obviously not hand made, but instead most likely generated by simulation profiles of these surfaces, using in-house data or from companies that run these kinds of surface simulations. I admit I do not know much about this kind of thing, and honestly this is part of the ‘secret sauce’ that developers are just not going to tell you about. Regardless, the surface characteristics show up in the data in various ways - damper speeds are different, suspension acts accordingly, the ‘chatter’ in slip ratios, differential behavior, grip levels and their associated friction circles - all of these behaviors are different for each of the surfaces and a big reason why you should listen to the damn pace notes.

Does any of this mean Dirt Rally’s tarmac physics is just gravel physics turned up? Looking at it from just the three components of vehicle dynamic modeling and their associated data, it is my opinion that no, it’s not just gravel physics turned up - the variables involved are too complicated for that to be the case. I hope what I have demonstrated here is that you cannot simply change a few numbers and voila, you have tarmac stages. Is the tarmac, then, just not given the same love as the gravel physics? Maybe, but again, many of the techniques you use to manage tarmac driving are transferable to Dirt Rally just fine. Grip, in a way, is a subjective term when used in the context of player feedback, because what constitutes correct grip level runs a wide gamut from one person to the next. You have players who love the physics of WRC9 but can’t stand DR2.0, and vice versa. How is that possible?

Which leads me to the last part to all this, and one that has as much to do with human psychology than physics - Immersion. All the physics in the world is not going to do you any good if your customers are not ready to believe it. Academics do not care about ray tracing or how accurate the car’s sounds are portrayed - but these elements, whether you realize it or not, play a massive role in your driving experience. It’s worth watching this video and seeing what the Dirt 5 developers have to say about immersion. 

Now, I do not believe this to mean that you need to throw out realism in order for a game to ‘feel’ real. But this aspect of how the player interacts with the game matters a lot, especially when it comes to control. Why do sim racers spend thousands for motion rigs and VR and direct drive fanatec wheels? They want that precious feedback, and the more they can get, the better. 

Consider the field of view for any sim racer. Developers can manipulate just this one variable to completely change the way that you sense speed in the game. Dirt Rally uses this to make the cars feel faster than you would with adjusted FoV. Setting up correct FoV, the game suddenly feels completely different - the cars ‘feel’ slower, and you get a stronger sense of the up and down motion of the car as well as jumps. It becomes a different experience. But why would the developers make the cars seem faster? My theory is it is a design choice to make the player feel that sense of danger that all of us human beings have when we experience high speeds. Higher feelings of danger leads to more adrenaline, which leads to heightened emotions, which results in the player feeling more engaged and ‘immersed’ with the game, and feeling like they are really racing, even though they are sitting down staring at a screen. The other part, consequently, is to compensate for that particular vehicle model quirk I mentioned earlier, about the car rotating on its axis.

Regarding this quirk, the game, as far as I know, only tracks two points on the car to determine its world location - the center and the front of the car - to determine the car’s yaw position. Brian Beckman provides the basis for determining car and tire position in his articles on simulation physics, “The Physics Of Racing Series” (P.104). Think about a game like beamng where the car is able to twist and warp around the chassis, and you’ll realize how restrictive just having two points can be. You essentially are driving a go-kart, but with imagined suspension systems. You have to hide this fact in various ways, and every game does it differently. RBR does an interesting thing with its audio, whereas as the car suspension rolls, the car audibly ‘creaks’, as if the suspension metal were bending back and forth. This is an excellent design choice and provides feedback that the car is experiencing weight transfer. What the limited car modeling means for Dirt Rally, is that the car seemingly develops lateral grip faster than it should. If you look at heli-cam replays you can see the car does not seem to move around naturally, but almost like a toy car in some ways. Hood cam/dash cam and the like, however, look more natural thanks to the game design. Remember, this is just a visual representation of the car’s behavior, it does not mean this is what’s actually happening under the hood, so to speak. If the driver feels like they are going fast, they will also be more likely to slow down more than they should, and thus, less likely to encounter ‘catastrophic’ oversteer or understeer. You can still be fast with the default FoV, it just means you need to overcome that initial feeling of fear to get more out of the car. Real race cars are just the same - feeling 2 G’s in a car for the first time is a terrifying experience, but given enough track time, you just start to feel it as another variable in your quest for maximizing grip.

This is just one aspect to a game’s overall design that impacts driver feedback. FFB is another big and controversial topic and is its own special craft. Managing a wide variety of controllers is another. I am sure there are plenty more to list but I will leave it there for now.

Having looked into the tarmac issue with some depth, I cannot state with total conviction the quality of the Dirt Rally’s tarmac physics. In my experience, just from working with the data and my own in-game driving, Codemasters has provided more than enough variables to simulate some version of real life tarmac physics that is predictable yet challenging to drive on. Behaviors that you would expect on real-life tarmac, setup changes, and track characteristics are all present to varying degrees. Improvements and refinements can always be made, of course, but I find it hard to outright dismiss the physics as being completely broken in any way. For those that do find severe issues with the physics, I would only ask that you make sure you are not locking your wheels up while braking before you hit that post button.

Sources:

Segers, Jorge. “Analysis Techniques for Racecar Data Acquisition, Second Edition”. 2014.

Abid, Samir. Interview with Damian Harty. https://www.yourdatadriven.com/ep6-chassis-setup-made-simple-with-ex-prodrive-engineer-damian-harty/

Et, al. Interview with Aris Vasilakos. https://www.yourdatadriven.com/ep14-how-you-can-get-the-most-from-sim-racing-with-assetto-corsas-aris-vasilakos/

Beckman, Brian. “Brian Beckman: The Physics in Games - Real Time Simulation Explained”. https://youtu.be/I8GQCZgCNw8
 

Edited by ManicKodo
  • Like 4
  • Thanks 2
Link to post
Share on other sites

@PJTierney, this is some incredible research and an extremely interesting read!

@ManicKodo, I compared my feeling/sensation in the car as a driver with what you wrote with your physics words.
And I was hoping you would clarify that what I'm experiencing is linked to what you wrote.
I lack any sort of specific knowledge when it comes to these things but I'm very interested in it and eager to learn.
 

  1. 9 hours ago, ManicKodo said:

    the simple point mass vehicle model (Segers 402)

    (cfr. point 😎
    I do feel there's a hint of pivoting on a central axis. Sometimes, it can be noticed when watching the replay.
    That has to do with this Segers 402 model, right? 
     
  2. 9 hours ago, ManicKodo said:

    aerodynamic downforce is not modeled

    (cfr. point 😎
    Is this the reason why we have that sideways motion while cornering when we actually still have full grip? 
    Looks like four wheel sliding in replays but whilst the car should actually hunker down and grip going through the corner?
    (The "gravel physics but grip turned up" comment people use as a way of describing it) 
     
    Or is that connected to the Segers 402 model (central pivot point)? Or both perhaps?
     
  3. 9 hours ago, ManicKodo said:
    Are those dips to nearly zero, moments where the tyres have absolute grip?
    When watching footage of the tyres of rally cars (or other race cars) when the set off in slow motion, you sometimes see that one or more of them stop rotating for a fraction of a second. That's because the tyre's tread gets stuck on or grabs hold of something beneath the surface, right? Is that what those dips to near 0 represent?

    Could that also be representing a slight lift off the throttle with the tyre tread being able to grab hold of something beneath the surface as a result of decreasing wheel rotation?   
     
  4. 9 hours ago, ManicKodo said:

    we can see that when we develop oversteer, we do not see the same correlated drop in grip

    Is this why the cars have more grip than they should and we can carry more speed when sliding around corners than we should on gravel?
    When really throwing the car into a corner on gravel carrying too much speed (pretty much ass first), somehow the car still has grip while in reality I should slide wide or off the road. 
     
  5. 9 hours ago, ManicKodo said:

    This constant back and forth is modeling what tires on gravel surfaces do, which is digging into the dirt and trying to find grippier surfaces underneath.

    This just reminds me of when I was discussing the way snow and soft tyres work in my "How to Master Monte Carlo" video. hehe
     
  6. 10 hours ago, ManicKodo said:

    Now, we have to consider that Dirt Rally’s surfaces are not laser-scanned but instead crafted by hand

    As far as I know, the stages are indeed not laser-scanned, as that would be extremely costly to do. One race circuit already absorbs a lot of budget to laserscan let alone a whole range of stage consisting of multiple surface types which are a lot longer than any race circuit (apart from the Nürburgring Nordschleife). And as you mentioned, there are also a lot more variables to take into account on loose surfaces such as gravel or snow/ice.
    I don't think any studio developing a rally title would be able to pull that off at this point in time.
    It would definitely be a gamechanger but I don't think any studio has the budget to do it.
  7. 10 hours ago, ManicKodo said:
    I guess this part in some way explains why my experience speed differently when driving and watching the replay (same camera view).
    When watching the replay, it always looks slower than what I remember experiencing whilst I was actually driving the stage. 
     
  8. 10 hours ago, ManicKodo said:

    Regarding this quirk, the game, as far as I know, only tracks two points on the car to determine its world location

    I started adding bits to my post after I read something interesting because otherwise I would just forget it halfway through the read.
    Now, I realise that points 1 and 2 are probably better suited for this section of your text.
    Reorganising the whole post would have been time consuming so I solved it this way.

    In addition to the comment in points 1 and 2...
    Would it be possible or better to go for a 4 point system (one point for each corner/wheel) or perhaps 5 (each corner/wheel + central point)?
    Any idea how much more costly or time consuming that would be for a development team? 
     
  9. 10 hours ago, ManicKodo said:

    Hood cam/dash cam and the like, however, look more natural thanks to the game design.

    That's indeed what I experience. When driving and using the onboard or bonnet cam, it feels a lot more realistic/natural compared to watching the replay from the TV or heli cam.

 

To wrap this up, I have one final question for you @ManicKodo.
What do you think, should be the main focus of the team developing the physics for the next DiRT Rally title?

My suggestion would be to focus on adding aerodynamics downforce and try to find a solution to move away from the 2 point system.
Being able to address those two elements would be a major step in the right direction, right?

And again, thanks for the extensive research and posting your findings.

Link to post
Share on other sites
Posted (edited)
9 hours ago, SRD_SimVansevenant said:

@PJTierney, this is some incredible research and an extremely interesting read!

@ManicKodo, I compared my feeling/sensation in the car as a driver with what you wrote with your physics words.
And I was hoping you would clarify that what I'm experiencing is linked to what you wrote.
I lack any sort of specific knowledge when it comes to these things but I'm very interested in it and eager to learn.
 

  1. (cfr. point 😎
    I do feel there's a hint of pivoting on a central axis. Sometimes, it can be noticed when watching the replay.
    That has to do with this Segers 402 model, right? 
     
  2. (cfr. point 😎
    Is this the reason why we have that sideways motion while cornering when we actually still have full grip? 
    Looks like four wheel sliding in replays but whilst the car should actually hunker down and grip going through the corner?
    (The "gravel physics but grip turned up" comment people use as a way of describing it) 
     
    Or is that connected to the Segers 402 model (central pivot point)? Or both perhaps?
     
  3. Are those dips to nearly zero, moments where the tyres have absolute grip?
    When watching footage of the tyres of rally cars (or other race cars) when the set off in slow motion, you sometimes see that one or more of them stop rotating for a fraction of a second. That's because the tyre's tread gets stuck on or grabs hold of something beneath the surface, right? Is that what those dips to near 0 represent?

    Could that also be representing a slight lift off the throttle with the tyre tread being able to grab hold of something beneath the surface as a result of decreasing wheel rotation?   
     
  4. Is this why the cars have more grip than they should and we can carry more speed when sliding around corners than we should on gravel?
    When really throwing the car into a corner on gravel carrying too much speed (pretty much ass first), somehow the car still has grip while in reality I should slide wide or off the road. 
     
  5. This just reminds me of when I was discussing the way snow and soft tyres work in my "How to Master Monte Carlo" video. hehe
     
  6. As far as I know, the stages are indeed not laser-scanned, as that would be extremely costly to do. One race circuit already absorbs a lot of budget to laserscan let alone a whole range of stage consisting of multiple surface types which are a lot longer than any race circuit (apart from the Nürburgring Nordschleife). And as you mentioned, there are also a lot more variables to take into account on loose surfaces such as gravel or snow/ice.
    I don't think any studio developing a rally title would be able to pull that off at this point in time.
    It would definitely be a gamechanger but I don't think any studio has the budget to do it.
  7. I guess this part in some way explains why my experience speed differently when driving and watching the replay (same camera view).
    When watching the replay, it always looks slower than what I remember experiencing whilst I was actually driving the stage. 
     
  8. I started adding bits to my post after I read something interesting because otherwise I would just forget it halfway through the read.
    Now, I realise that points 1 and 2 are probably better suited for this section of your text.
    Reorganising the whole post would have been time consuming so I solved it this way.

    In addition to the comment in points 1 and 2...
    Would it be possible or better to go for a 4 point system (one point for each corner/wheel) or perhaps 5 (each corner/wheel + central point)?
    Any idea how much more costly or time consuming that would be for a development team? 
     
  9. That's indeed what I experience. When driving and using the onboard or bonnet cam, it feels a lot more realistic/natural compared to watching the replay from the TV or heli cam.

 

To wrap this up, I have one final question for you @ManicKodo.
What do you think, should be the main focus of the team developing the physics for the next DiRT Rally title?

My suggestion would be to focus on adding aerodynamics downforce and try to find a solution to move away from the 2 point system.
Being able to address those two elements would be a major step in the right direction, right?

And again, thanks for the extensive research and posting your findings.

Appreciate all your questions! They are all very good ones.

1. To clarify, Segers did not come up with this model, just that he talks about it in his book. I try to demonstrate that with the data and the visual thing with the heli-cam and BayouRally's comments which are spot on, imo.

2. Y'know, I'm not entirely sure about that one. It's a good theory if it is related to the simple mass point model. For me, I was looking mainly at the data for the aero differences, not so much the actual visuals. The immersion/visual thing is a whole other bag of problems related to control theory, of which I have very little understanding of, but we as gamers understand it intuitively in a way, thanks to all that controller/wheel practice.

3. Those dips are gear changes, so nothing all that special. The dips just mean that the rear tires in this case have become free-rolling wheels like the front wheels and are spinning (on average) at the same rate.

4. I used this particular sequence with the hairpin to just illustrate that slip angles work differently from tarmac, but there are definitely moments where you can oversteer too much on gravel. I've experienced the consequences of that plenty of times on the downhill sections in Greece! The attitude velocity/lat G graphs I show later give a better picture of what's happening, I think. While on gravel there is a wider operating band particularly in the oversteer region, you can still see that maximum lateral grip still funnels towards less oversteer, just to a lesser extent than on tarmac. It's a little bit tricky to demonstrate without video, because what does it mean to exceed 30 deg/s? For that matter, what does it mean to exceed peak slip angle? For me, it's more about at what attitude velocity do I suddenly start using corrective driving inputs to fight the oversteer/understeer that I am experiencing? That's really where I start to measure when I think it is too much/not enough rotation. Every driver will be different, of course. So this is all to say that I do think there's a limit to how much grip you can develop from drifting, particularly when your ground speed is quite high or your weight transfer management is way off.

5. I am relieved that my observations can be confirmed in this way!

6. It seems impossible, but I say never say never. Who knows what sim racing will be like in 10 years? 20 years? The future is bright.

8. This is a hard one, because in-game suspension changes on each tire are tracked, to the point I can tell you in degrees approximately how much the car rolls left to right on the front side and rear side, and how many degrees the car pitches front to rear (I hope to look into this with a future post, as it needs a lot of explanation). Those two points are used to represent the car's position visually and I am not certain how the game translates the vehicle and tire model into what we actually see on screen. This is some game design stuff that I don't have any experience with, really. All I can say is once you start getting to that point, of having the underlying physics really be represented on screen accurately, you are talking about soft-body physics like beamng.

As far as where the game should go from here, I of course have opinions about that!

  1. Definitely add at least one or more of these things: aero physics, or tire temperature effects, or tire pressure, with a more detailed tire degradation model than the current one. These should be de facto features for modern sim racers as that is where all the other sims are going these days, not to mention they are big major components to vehicle dynamics. If I'm right, these things were already in the works for DR2.0, so hopefully they can add them for the next title.
  2. More dynamic gravel physics! Basically just add more features. Varying levels of road softness, mud and sand surfaces, random rocks that need to be avoided - just some level of craziness to off-roading! 
  3. More dynamic tarmac! Same idea - cracks and pebble surfaces, newly paved vs. old pavement, more hazards. Would love more of these kinds of things to really throw off the player. The success of monte carlo player experiences should suggest that players like those kinds of environments. 

As far as having a better model than the single point mass model, I actually think it is ok for now? Going beyond that just seems like a step too far at the moment, if beamng was any indication, you really need to wait a while longer for the technology to get there, either through optimizations through iteration or pure computer horsepower. The three points above I think would provide plenty of variables to justify in a sequel. I would love to explore these issues some more in games like ACC as they are very friendly to data logging (up to 200hz sample rate! Yowza!). But in the meantime I think on the physics side of things, those three things would make the DR experience much more interesting.

Edited by ManicKodo
Link to post
Share on other sites

@ManicKodo Thanks for the reaction, mate. Appreciate it!

Looking forward to hear your thoughts when the next DiRT Rally comes out at some point in the somewhat near future (fingers crossed!). 
You know, comparing DR3 with DR2.0.

Btw, have you ever thought of comparing DR2.0 with DR? Some people mentioned they prefer the DR physics over DR2.0 and perceive it as more realistic.

Link to post
Share on other sites
17 hours ago, SRD_SimVansevenant said:

@ManicKodo Thanks for the reaction, mate. Appreciate it!

Looking forward to hear your thoughts when the next DiRT Rally comes out at some point in the somewhat near future (fingers crossed!). 
You know, comparing DR3 with DR2.0.

Btw, have you ever thought of comparing DR2.0 with DR? Some people mentioned they prefer the DR physics over DR2.0 and perceive it as more realistic.

I hope it comes soon! I don't think I'm going to do any comparisons with DR1, I just don't have that much interest as everything seems to be focused on DR2.0 and beyond.

Currently learning to play AC with controller to get a feel for their tarmac physics, very different experience!

  • Thanks 1
Link to post
Share on other sites
Quote

Is this why the cars have more grip than they should and we can carry more speed when sliding around corners than we should on gravel?
When really throwing the car into a corner on gravel carrying too much speed (pretty much ass first), somehow the car still has grip while in reality I should slide wide or off the road. 

 

On gravel, I find that quite realistic. I think the tires don't lose grip when slipping and they gain something when spinning.
But the fact that it works similarly well on asphalt bothers me.
If I oversleep a turn, I always try to get the car into the turn by sliding sideways. The braking effect is enormous. (In Assetto Corsa this is not so easy).

There is another problem with FWDs on tarmac. If I enter a corner a bit too fast and the car is already understeering, I can still accelerate a bit and this tends to improve the situation, although my feeling tells me that it should rather get worse, because the tires are already overloaded.

Yes, and the car always turns a bit around the center, the rear wheels slide out. That may be realistic on gravel, but on asphalt the other racing games don't have that either.

 

Translated with www.DeepL.com/Translator (free version)

Link to post
Share on other sites
Posted (edited)

I've been doing a lot of work with ACC lately (truly impressive and unbelievable game!) and it makes the differences in physics very apparent. It is very obvious how much more detailed the handling is, the effects of aero, how much more unsettling bumps and kerbs can be, and what laser scanned tracks bring to the table. I really like the effects of understeer, how it is achieved, and how to resolve it - it makes sense the way you would want it to. I think DR2.0 lacks this kind of fidelity. Then again, the GT3 cars are a completely different beast and setup, but it's really about the details - it's like the difference between a 128kbps mp3 and 16-bit FLAC.

The game is crazy hard though - not a game for casuals at all. But what a treat for hardcore sim racers!

It's interesting though, that DR2.0 has more logged data channels than ACC, even though ACC developers make a point to offer data analysis opportunities. I wonder if there are similar restrictions in GT3 regarding data logging?

Edited by ManicKodo
Link to post
Share on other sites

This might well be the first time that I see a forum post with quotations... Proper job here @ManicKodo!

I don't know much about physics design in games, but I have this feeling that the issues the DiRT Rally are inherent to its "simplified" two-point vehicle simulation, but not because it's an inherently bad system. 

As you said, it has been used in many racing games over the years and good developers, with some "art", have managed to get very convincing car handling out of it. 

The problem in the DiRT series is that it has to deal with two very different surfaces, tarmac and gravel (or otherwise loose surface), while using the same simplified car behavior model. The "tricks" used to make the car feel good on one surface become problems when they are met with the completely different behavior the other surface induces. 

Historically, the DiRT series has focused on gravel handling and the skill of the developers has created one of the best simulations of this surface in gaming history (if not The Best), but the tweaks made to reach that level make the cars on tarmac feel not at the same level. 

Of course this is just my speculation, I'm very interested to hear your opinion on the matter! 

  • Like 1
Link to post
Share on other sites

Thank you! It's very hard to say for me, what the main issue is. I think a good place to start off is looking at the sheer difficulty of simulating the physics of a given rally stage. The suspension is heavy duty and extremely active, there are random elements all about the road surface, the car needs to be able to handle tarmac and gravel, temperatures from icy cold to blazing hot, tires that are super thick and skinny but tough enough to withstand the rocky surfaces, day and night stages, long endurance weekends, AND having a codriver throwing constant pace notes into your ear? Oh, and you fly into the air on occasion. 

It's almost a fool's errand to try and simulate all of this while still providing a worthy gaming experience. I see how much work Kunos has put into Assetto Corsa Competizione, and that's a very specific discipline that they are modeling! To try and add this whole other experience on completely different surfaces, well, it would require some serious ambition on the part of the developers. I don't know enough about game development to know if such a thing is possible or makes financial sense. And to be perfectly honest, I don't know exactly what compromises were made between the gravel and tarmac physics! It's my belief that where the technology is going these days, it's not just about tweaking the engine until the numbers look right - there needs to be some added depth to the models (especially the track model) to really hit that next level. Whether it's moving away from the single point model or adding greater variables and gradients in the tire models, or really pushing the track model in ways that have not been attempted yet - any one of these things could make a significant leap in rally games. I would love to see any of these things happen! Not that it would be a bad game if they kept the physics and just added a bunch of new stages or a robust career mode, but for me personally I would love to see some bigger advancements in the physics.

  • Thanks 1
Link to post
Share on other sites
On 6/25/2021 at 3:20 AM, dgeesi0 said:

less is more. what people do is add more things to go wrong or feel right. be simple. it will work.

The thing is, a car interacting with the road surface is not a simple thing. There are so many variables and so many different complex concepts at play that I don't think a simple model would be able to represent everything that happens on a rally stage realistically. 

Take, for example, how both the tyres and the road surface deform in the different phases of the corner... 

  • Like 1
Link to post
Share on other sites
On 6/26/2021 at 11:44 PM, carpa said:

The thing is, a car interacting with the road surface is not a simple thing. There are so many variables and so many different complex concepts at play that I don't think a simple model would be able to represent everything that happens on a rally stage realistically. 

Take, for example, how both the tyres and the road surface deform in the different phases of the corner... 

the funny thing is intelligence teaches exactly the total opposite of what you just said when actually feeling driving. so many variables.... Q whiteboard presentation of what is set to be achieved during driving simulations ....multiple maths equations for nothing because they trying to apply math to a actual feeling.......woah too much. 😁

 

when you break down to the lowest factor there isnt many varibles needed for a basic good feeling of driving. this is what puzzles me.

 

how do you do this codemasters EA send me a big check and i will tell you how. its actually really simple. 😁 ( im not even joking ) ching ching

 

Edited by dgeesi0
  • Agree 1
Link to post
Share on other sites
6 hours ago, dgeesi0 said:

the funny thing is intelligence teaches exactly the total opposite of what you just said when actually feeling driving. so many variables.... Q whiteboard presentation of what is set to be achieved during driving simulations ....multiple maths equations for nothing because they trying to apply math to a actual feeling.......woah too much. 😁

 

when you break down to the lowest factor there isnt many varibles needed for a basic good feeling of driving. this is what puzzles me.

 

how do you do this codemasters EA send me a big check and i will tell you how. its actually really simple. 😁 ( im not even joking ) ching ching

 

In some ways, I have to agree: no matter how complex and in-depth the model is, in order to achieve a good feeling of driving the car some "art" is still needed from the developers to make things feel right. 

Codies know this very well and this is also why they have Jon driving the game and giving feedback to make the cars handle right. 

I still think that simulating many variables and aspects of driving helps achieve the higher levels of realism, but this is only true if the developers know how to tweak things in their model in order to achieve a realistic behavior from the cars. 

If the model is too complex for them (and I think DiRT developers are too good for this to be the case) they'd be better off with a simpler model that they can fully use. 

  • Like 1
Link to post
Share on other sites
  • Social Media & Community Team
10 hours ago, dgeesi0 said:

how do you do this codemasters EA send me a big check and i will tell you how. its actually really simple. 😁 ( im not even joking ) ching ching

https://codemasters.csod.com/ux/ats/careersite/2/home/requisition/193?c=codemasters

 

There's your cheque 😉 

Start working on F1 and then move over to rally 😄 

  • Haha 2
Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...