GT5 physics model - how complex is it?

[Fredde_swe]

I assume you guys do not use traction control. I don't either.
If you step on the gas even a small amount then let off the tire indicator turns red, right? Then it would make sense the next time you get on it it it would provide better traction assuming the tires are new. Take a look at any staging area for drag racing.

In this game, the tire heats up and always overheats.

The you are pushing your car to hard

 

[Jarmo]

In this game, the tire heats up and always overheats.

As far as I know the tire indicator shows the surface temperature. The surface temperature can overheat very quickly (also cools off quickly) but when you race for a prolonged time the whole tire gets warmed up and this increases the grip.

 

[ZohsixGT5]

One of the things I've thought for a long time is that PD has come up with a solution that makes driving the game such a unique experience. I think they exaggerate many aspects of car dynamics specifically to make the experience feel more "real" with the limited feedback mechanisms at our disposal. If you've played many other SIMS, one thing you'll notice is that in PD's products the cars feel extremely unique. You will not, even for a second, think that your RUF Yellowbird feel awfully similar to your RUF RGT. Even though they are evolutions of the same car, and even though they both feel very RR, the are also quite different. In every other SIM I've ever played, the cars feel less unique than in the GT series, which I think is key in the attraction GT has in appealing to "car enthusiasts" and not just "motor-sport enthusiasts". The thing is, all PD has to communicate these differences are sight, sound, and FFB in a steering wheel. This isn't even close to the amount of feedback a real car gives. So, how do you communicate the uniqueness of each car? How do you communicate how drafting works (for example)? Well, you exaggerate it sufficiently that the person using your product feel the difference keenly. You make cars that are neutral very neutral, you make cars with oversteer have lots of oversteer, you make cars with a soft suspension very soft and cars with a stiff suspension very stiff, you make drafting give more of a boost in speed than it should. Thus you get the "feel" of the real car through the limited sensations available for PD to work with.

Personally, this is a method that "really works for me" because that's exactly the experience I'm looking for: I want to know "what does driving an AC Cobra feel like? What does driving a Lancia Stratos feel like? etc. I don't know how complex it is, but however PD does it, it is very convincing.

Well said.

 

[Exorcet]

One of the things I've thought for a long time is that PD has come up with a solution that makes driving the game such a unique experience. I think they exaggerate many aspects of car dynamics specifically to make the experience feel more "real" with the limited feedback mechanisms at our disposal.

I feel that it does the exact opposite. ESPECIALLY with the aerodynamics. The suspension settings seem very suppressed too, I'd expect much bigger differences when I go from a max setting to a min setting, etc.

If you've played many other SIMS, one thing you'll notice is that in PD's products the cars feel extremely unique. You will not, even for a second, think that your RUF Yellowbird feel awfully similar to your RUF RGT.

I don't agree with this either. I don't think GT captures all the subtle differences between cars very well. Sometimes, not even the huge differences. The Viper ACR in game is just an 08 SRT-10 for example. Don't get me wrong, I can tell one car from another. The 458 drives very differently from a Viper, and both are nothing like the GT-R, but they're all fundamentally different cars. On the other hand, take a Viper and Corvette, or even the Z06 and ZR1 and compare them, and they seem a little too similar. Every review I've read pitting the Viper vs Corvette puts them worlds apart. Like night and day. It's not that way in GT.

 

[DJBLITZKRIEG]

Why come you can't use proper grammer?

Hahahaha

 

[ugabugaz]

I feel that it does the exact opposite. ESPECIALLY with the aerodynamics. The suspension settings seem very suppressed too, I'd expect much bigger differences when I go from a max setting to a min setting, etc.


I don't agree with this either. I don't think GT captures all the subtle differences between cars very well. Sometimes, not even the huge differences. The Viper ACR in game is just an 08 SRT-10 for example. Don't get me wrong, I can tell one car from another. The 458 drives very differently from a Viper, and both are nothing like the GT-R, but they're all fundamentally different cars. On the other hand, take a Viper and Corvette, or even the Z06 and ZR1 and compare them, and they seem a little too similar. Every review I've read pitting the Viper vs Corvette puts them worlds apart. Like night and day. It's not that way in GT.

I don't feel the way you do about the Vette and Viper. Each car type feels different to me, even the Z06 vs ZR1 feel different.

 

[HDHSVI]

"On topic - I don't know but it's amazing.
And it's not "just numbers", it's much more than that.
I mean, for example - Zonda feels twitchy as it should, 458 has that bit of high speed slide, McLaren F1 is perfect (again, as in real life), not to mention GT-R, 8C Competizione is horrible and has all the attributes of a real life 8C (delay in transmission, bad handling etc), MP4-12C apparently feels just as it should in real life.
Some car's aren't done good (for example Lancer Evo X) but 95% of the cars are spot on.
Some things aren't done great (such as tyre physics, aerodynamics) but overal it's great and by far the best on consoles. "


Couldn't agree with you more! - one more to add to the not near acurate list the Shelby GT350R.

Cya,
Hankster

 

[oppositelock]

I don't know about the physics engine, but the apparently completely random weight distribution numbers makes me doubt the sophistication of the vehicle parameters that are put into it.

 

[Dillinger812000]

Why come you can't use proper grammer?

Why come you can't use correct spelling?

 

[CSLACR]

I feel that it does the exact opposite. ESPECIALLY with the aerodynamics. The suspension settings seem very suppressed too, I'd expect much bigger differences when I go from a max setting to a min setting, etc.


I don't agree with this either. I don't think GT captures all the subtle differences between cars very well. Sometimes, not even the huge differences. The Viper ACR in game is just an 08 SRT-10 for example. Don't get me wrong, I can tell one car from another. The 458 drives very differently from a Viper, and both are nothing like the GT-R, but they're all fundamentally different cars. On the other hand, take a Viper and Corvette, or even the Z06 and ZR1 and compare them, and they seem a little too similar. Every review I've read pitting the Viper vs Corvette puts them worlds apart. Like night and day. It's not that way in GT.

I agree with your point as a whole, but from everything I've seen, the Viper's and Vette's should be more like each other than most any other cars in the game. (redundant sub models not included)

I don't think GT5's physics are very complex compared to reality, no. I'm not really sure how individual aspects line up with other games, but the "package" in GT5, while flawed in most if not every area, still tallies up to pretty realistic IMO.
For example: Ride height - higher equals better traction while accelerating, lower equals faster turning.

Downforce - Lower doesn't increase top speed or any acceleration barely at all. Higher downforce also equals accelerating traction - ***From a dead stop.***

I don't think forward motion while spinning tires is enough, which results in breaking the tires loose causing to much time loss. This is particularly noticeable in stock FWD cars that can't go full throttle in first gear without smoking the rubber off the rim.

Regardless of what people have told me, from what I know, and now a book I've seen parts of written by Carroll Shelby only further confirms my original thoughts, most suspension tuning is backwards, meaning the whole left side. Ride height, spring rate, dampers, and sway bars.
The prime example is of course, that stiffening the rear equals understeer in GT5, no if's and's or but's about it.

And lastly for now, it seems there is a "cornering scale" somewhere in PD's basement that's been there for at least a decade. What it means is different cars have different levels of grip, I believe there's 3-5 "levels" of grip, ranging from complete crap (Aston Martin V8 Vantage) others fall into "inbetween" categories, and cars like the GTR, NSX, etc have "top grip" meaning other regular cars of similar specs will just always corner worse, regardless of tuning suspension.
To add to that, there's "lightweight" grip, "suspension" grip, and possibly "chassis" grip, in which, you will find - good "suspension" (cars known for having good suspension from the factory) "Suspension" cars will take certain corners on each track the fastest, while "lightweights" will take other corners fastest, etc.
A perfect example is the NSX and F430 Scuderia, I noticed these patterns in GT4, and rediscovered them in GT5, in the example, you will find the NSX and F430 can take each corner at the Nurburgring at nearly the same speed, and it's found measuring speeds mid corner, these are "suspension" cars, so the corners an Elise can take best will usually be different, although they're all three fully credited "grip" cars with the highest levels of cornering grip from street cars available in the game.
Oh, and there's "downforce" cars too, the ones with hidden downforce which receive extra traction off the line, and through most corners just because PD said "here, have some downforce". All of them would be fine, except there's nothing to improve the cars credited "bad" anything from the start. (Meaning a poor cornering car can never be made to corner a "level of grip" higher, aka - A "level 1" (lowest) can never compete with a level "2", and level 2 can't be made to equal "level 3", etc.

And that's my take on the simplicity and I believe some stupidity in GT5's physics.

 

[GTP_Gearhead]

Simple maths?

OK, read this......

http://phors.locost7.info/contents.htm


....and come back with a straight face and tell me its simple maths. The physics around the likes of tyre dynamics alone are hideously involved.

You may be getting a pop quiz as well for displaying such a huge level of understatement and over confidence.


Scaff

OMG. Just looking at section 8 has me ready for a drink.

Computing a cars velocity
"How fast it is going in x direction, ignoring its motion in the y and z directions."
The velocity of the race car formula.
race-car-vx car161
race-car-vy car161
race-car-vz car161
define vx(1(-x2x1)(-t2t1))

This stuff is insane.

 

[plate88]

I've done some searching and guess what? Found nothing on GT5 physics, or at least the workings behind what drives it. I'd imagine they're fairly complex, but probably nothing like a pc sim.

What concerns me with GT5 is that we have no real data like telementry to tune our cars properly. Not to mention some of the settings are backwards. Be nice to have tire temps, ride heights, suspension travel, etc. Yes they do have some sort of tire temp data in there by looking at the HUD, but to me thats a pretty vague way of doing it. Better than nothing I suppose. Hope PD will put telementry into maybe GT6. I would enjoy tuning a car alot more and trying to decipher all the data to optimize the cars capabilities for tracks would be fun.

I do like GT5's physics alot, but some things in them just don't make sense.

I'd like to add, I believe the physics are best when using comfort softs or sports hard. I feel anything more than racing hard is overkill and racing softs tilt GT5 into the arcade section for me. Just my opinion, but I see alot of lobbies online with racing softs and hope people's view on the physics isn't from that standpoint.

 

[brainfade]

Every physics model is very very complex, almost mind boggling to the average guy. The only 'hole' so to speak in the physics model is the tyre physics regarding physical deformation and the tyre's dynamic contact patch with the surface of the track, possibly the surface abrasiveness of the track's effect on the tyre < remember with some cars, especially race cars, 10-50% of the suspension travel can be the tyre itself, hence why cars appear to 'bounce' in real life, the cars in GT5 and 95% of most sims appear to be stuck rigid.

Few examples of the 'bouncing' and tyre behaviour that I'm referring to (ignore the sh***y music):

 

[CSLACR]

Every physics model is very very complex, almost mind boggling to the average guy. The only 'hole' so to speak in the physics model is the tyre physics regarding physical deformation and the tyre's dynamic contact patch with the surface of the track, possibly the surface abrasiveness of the track's effect on the tyre < remember with some cars, especially race cars, 10-50% of the suspension travel can be the tyre itself, hence why cars appear to 'bounce' in real life, the cars in GT5 and 95% of most sims appear to be stuck rigid.

Few examples of the 'bouncing' and tyre behaviour that I'm referring to (ignore the sh***y music):

I wrote a post showing quite a few holes, I disagree tire physics are the only "hole" in GT5's physics.
How about ABS? (Something I forgot in my post)

 

[Bushby_23]

From what I read in a magazine interview with kaz, its alot like a giant spreadsheet. Anything that can be represented by a value for each individual component is. All these different values are put through many equations with other values that represent different things on the car. The number that is the result of the equation is then put into the model and then that will be how the car handles and drives.

 

[Sohvakettu]

I wrote a post showing quite a few holes, I disagree tire physics are the only "hole" in GT5's physics.
How about ABS? (Something I forgot in my post)

Tiremodel doesn't have the only holes in the game but the effects of braking for example depends on how the tire model is working. Almost everything the car does it does through it's tires so tiremodel is the foundation for everything else.

 

[CSLACR]

Tiremodel doesn't have the only holes in the game but the effects of braking for example depends on how the tire model is working. Almost everything the car does it does through it's tires so tiremodel is the foundation for everything else.

True but you called it "the only hole", to which it is actually a rather enveloping problem.
Also, ABS can easily be modeled without a proper tire model, it's braking that's "hindered" by the tire model, not how the brakes apply, but rather, the slowing motions of the car that would be influenced.

 

[sinbad]

There's also the way that light cars have too little braking and cornering advantage over heavier cars. Surely an indicator that something is amiss.

 

[DaveM-sport]

I won't start listing here all shortcomings in GT5 physics, but due to the sheer number of cars in GT there's one thing in particular that probably will never get as advanced as pc sims simulating just a few cars: the suspension model. What PD probably does is simulating in a very general way just the general differences between different suspension types (McPherson, live axles, semi-indipendent axles, double wishbones), without getting into specific details (suspension travel and preload, exact linkage points and kinematics, panhard rod lenght, etc).

They also don't appear to use proper camber and toe values for every car, although this is relatively easily available data.

fully agree although there's alot of stuff thats very close or identical to real cars such as gear ratios of each gear and final drives and suspension setup in terms of spring rates and camber is reasonably close on most cars I've looked at.

Although Having rear camber and toe adjustment on an AE86 is alittle silly in fairness

 

[robbo6]

Don't forget GTs brake balance has always been wrong IMO. Brake force and Balance should be two separate variable adjustable 1% at a time. This 1-10 front and rear balance strength thing is bunk. I want to set a car up with 56% front 44% rear balance. It's impossible to do that in GT5 (and previous GTs if memory serves).

 

[alex291190]

[REARLEFT]
BumpTravel=-0.010
ReboundTravel=-0.200
BumpStopSpring=70000.0
BumpStopRisingSpring=2.00e7
BumpStopDamper=2500.0
BumpStopRisingDamper=7.00e5
BumpStage2=0.090
ReboundStage2=-0.090
FrictionTorque=1.50
SpinInertia=0.8 
PushrodSpindle=(-0.120, -0.080, 0.000)
PushrodBody=(-0.250, 0.240, 0.000)
CamberRange=(-2.5, 0.1, 21)
CamberSetting=10
PressureRange=(140.0, 1.0, 51)
PressureSetting=20
PackerRange=(0.000, 0.0, 0)
PackerSetting=0
SpringMult=.90
SpringRange=(25000.0, 1000.0, 11)
SpringSetting=5
RideHeightRange=(0.060, 0.001, 21)
RideHeightSetting=0
DamperMult=1.0
SlowBumpRange=(2800.0, 100.0, 10)
SlowBumpSetting=4
FastBumpRange=(1700.0, 100.0, 10)
FastBumpSetting=4
SlowReboundRange=(4700.0, 100.0, 10)
SlowReboundSetting=4
FastReboundRange=(3000.0, 100.0, 10)
FastReboundSetting=4
BrakeDiscRange=(0.030, 0.0, 1)
BrakeDiscSetting=0
BrakePadRange=(0, 1, 5)             // pad type (not implemented)
BrakePadSetting=2
BrakeDiscInertia=0.715              // inertia per meter of thickness
BrakeOptimumTemp=550.0              // optimum brake temperature in Celsius (peak brake grip)
BrakeFadeRange=1100.0               // temperature outside of optimum that brake grip drops to half (too hot or too cold)
BrakeWearRate=4.000e-011            // meters of wear per second at optimum temperature
BrakeFailure=(1.43e-002,5.28e-004)  // average and variation in disc thickness at failure
BrakeTorque=1400.0                  // maximum brake torque at zero wear and optimum temp
BrakeHeating=0.00195                // heat added linearly with brake torque times wheel speed (at max disc thickness)
BrakeCooling=(3.000e-002,4.000e-004)  // minimum brake cooling rate (base and per unit velocity) (at max disc thickness)
BrakeDuctCooling=1.001e-004         // brake cooling rate per brake duct setting (at max disc thickness)

That's from GTR2 - which uses the GMotor2 physics engine, and that's only the rear left corner. - just to give you an idea of how complex a physics engine can be

 

[shirakawaa]

Don't forget GTs brake balance has always been wrong IMO. Brake force and Balance should be two separate variable adjustable 1% at a time. This 1-10 front and rear balance strength thing is bunk. I want to set a car up with 56% front 44% rear balance. It's impossible to do that in GT5 (and previous GTs if memory serves).

Quite true! It would be easier to use and understand that way.
However this isn't really a "physics" issue, but more like UI/strange choices.

 

[HKS racer]

Every physics model is very very complex, almost mind boggling to the average guy. The only 'hole' so to speak in the physics model is the tyre physics regarding physical deformation and the tyre's dynamic contact patch with the surface of the track, possibly the surface abrasiveness of the track's effect on the tyre < remember with some cars, especially race cars, 10-50% of the suspension travel can be the tyre itself, hence why cars appear to 'bounce' in real life, the cars in GT5 and 95% of most sims appear to be stuck rigid.

Exactly, tyres deformation bouncing is not yet properly modelled in actual sim racing games. Calculate all that stuff in realtime toghether with all the other parameters and modern graphics is a LOT of work for a pc even more for a console. rFactor 2 some months ago has shown an early WIP stage of their brand new tyre wear and deformation model, obviously I can't judge from a video, but it seems a (big? little?) step in the right direction. rF2 will dedicate a core only for the physics, and the other core for graphics and stuff.

Sure, lack of tyre deformation is not the only short coming in GT5 physics but I think the ABS issue is more a lazy / rushed issue not a physics engine limitation. ABS off doesn't work very good, tyres tend to lock too early and too often but Iif they want to fix it they can do it quite easily.
In the end I think GT5 physics is good enough for a console racer, expecially without using racing tyres.

 

[Nikked]

Are the days of just enjoying a game without worrying about tire contact patch physics engine over?

 

[alonsof1fan91]

OMG. Just looking at section 8 has me ready for a drink.

Computing a cars velocity
"How fast it is going in x direction, ignoring its motion in the y and z directions."
The velocity of the race car formula.
race-car-vx car161
race-car-vy car161
race-car-vz car161
define vx(1(-x2x1)(-t2t1))

This stuff is insane.

Well, Im almost done with studying mechanical engineering, so I can understand some of the equations in Scaffs post, but alot of them are new to me. But not everything is out of this world.
F.e. I don`t find it difficult to read the script which is posted above me (the GTR2 one). Don`t forget one thing, those people who write those engines are just mere humans like you and me.
They probably aren`t even engineers, I guess most of them are working in the IT buisness.

 

[HKS racer]

Are the days of just enjoying a game without worrying about tire contact patch physics engine over?

Well no, you are right, but whey they'll add that thing properly we'll enjoy racing sims even more. 👍

 

[tomhart9]

Poly' need to get hold of this program called "Adams". It's incredibly realistic simulation software, I was shown a demo of it at a uni open day I went to recently. They use it for simulation of race car physics to make it easy to test things like suspension much quicker than having to go out and actually do it.
Obviously they run on top end computers so it's doubtful a PS3 could handle it, but maybe a future generation console could handle something similar....

 

[silicon1138]

Are the days of just enjoying a game without worrying about tire contact patch physics engine over?

Not at all. This is a forum for discussion of all things GT5. I like the game so much that it makes me think deeply about what's involved in its creation, i also enjoy reading GTP users comments about such things. Thanks for all your comments so far by the way.
Kaz seems to answer the odd question that GTP users send him - maybe someone could ask him just how many components his car physics model employs 💡

 

[Lambob]

How many Teraflops does the PS3 Cell processor handle, does anyone know? well, pluck that value, with all the car variables at your disposal, and steering input, minus graphic rendering, minus kernel overlay, minus sounds, minus notifications of low battery on controller 1, and the answer I think lies in hundreds of thousands of possible behavioral combinations the car will be at that point in time on the track, per second.

Throw in a wrench for "physics bugs, (especially online) where they are also different in every room you join, unfortunately.

Looking forward to spec 2.0 and it's 10 patches after that.

 

[TBR 427]

With regards to the tyre deformation, in games it's usually a case of if you see it, you think you feel it, not, I can see it therefore it exists.

The colliders in-game are STILL either polygonal (they can deform, but are made up of flat, triangular sections) or raycast (sort of like a laser security system).

While you might SEE a wheel in a game deform and blob around a bit, basing the physics off that resulting mesh would be fairly stupid - best case scenario your car handles like it has 40+ flatspots on each tyre. What's actually happening is a curve or single point is being mathematically defined around the wheels at different points (inside, mid, outside) at their bottom-most contact point/s. THOSE are what drive the physics for the car.

Once that is solved, the resulting transform/rotate is applied for the next update cycle (500Hz for GT5, if I remember). After ~10 updates, it draws the frame, renders to screen the current position of the vehicle, etc bla bla bla usual stuff.
If it's got VISIBLY deforming tyres, THEN it calculates that, based on the difference in the cars position compared to the last frame (not update). Soft body is bloody hard to deal with, considering it doesn't contribute to the physics (too slow, nowhere near accurate enough) and you need the tiny frame or two of lag to make it look realistic. That deformation is rendered in the next frame.