Dynamic weather and track conditions?

[Stonemonkey]

Regardless of whether or not this technique is used in GT5, there will be some system for reducing the level of detail as the cars go into the distance and it could be that the maximum level of detail is only reached in close ups of the car when not much else is visible onscreen so it might not be a valid argument that the resources to process a 500k poly car should be used elsewhere.

 

[Kryz11]

Or when Rain comes over, we might get a Framerate between 30-60 FPS.. My guess around 45, so PD has more time to improve rain effects.

 

[Stonemonkey]

Or when Rain comes over, we might get a Framerate between 30-60 FPS.. My guess around 45, so PD has more time to improve rain effects.

And you base your guess on?

 

[Dravonic]

Or when Rain comes over, we might get a Framerate between 30-60 FPS.. My guess around 45, so PD has more time to improve rain effects.

I don't know. Kazunori is really obsessed with 60fps. He even went as far as saying that a game can't be considered a game if it's not running in 60fps.

 

[vigi147]

I was just watching a WTCC race in Curitiba, Brazil and started thinking..
It was an interesting first race ( weather-wise ). At the start of the race it was pouring rain, very wet, puddles everywhere so most of the teams decided to go for the wet tires. About 5 laps into the race it stopped raining, the sun came out and the track started to dry. At lap 10 it was already almost completely dry and it was very interesting to see drivers struggling with the tires and desperately trying to find wet parts of the track in order to cool the tires off a little bit - and then flying out in the first corner because they didnt. This got me thinking .. will we see something like this in GT?

I know weather is in, but it's probably not going to be dynamic to this degree. But i would love to see this in the game someday (GT6 maybe?). Being able to choose slicks or wet tires according to the weather forecast and then struggling to cool down the tires when the forecast fails and the track suddenly dries. Looking for puddles just to kick down the temperature a little bit and get better traction..

Hope there's not a similar thread .. if there is one, i apologise

 

[daymobrown]

For me, weather needs to be digital or fully analogue, if the track will slowly dry off it needs to develop a dry line first or there's no point.

 

[analog]

Weather fully analogue only? There's a reason i have my PS3 indoors...

 

[hampus_dh]

GT5P had 200k polygon cars while GT5 cars will have 500k.

Source?

 

[Griffith500]

VISUAL ELEMENT

I'd totally forgotten about the wet tracks in GT3 and GT4.

The SSR5 wet in GT3 was time-trial only, if I remember correctly. Anyway, it looked superb, but there was no precipitation, and I assume (can't remember) the grip level was constant over the entire track.

In GT4 we got a series of "special conditions" races at Tsukuba where you raced against one opponent. Again no precipitation, and I don't recall a difference in grip, although I may have been convinced of it at the time (this was probably due to my inconsistent driving style and eagerness to blame the game...)

So, the visual element has been working for some time, but has obviously incurred quite a performance penalty on the old hardware. The "shaders" etc. will need recompiling / re-optimising / re-writing for the new hardware, but it's essentially done.
What's missing is precipitation, something that is very hard to do convincingly without faking it with a lot of fogging (wait, that's all it is, right? Heavy, falling fog )

PHYSICAL ELEMENT (a hypothetical rambling...)

Then of course, the modeling of the physical aspect on the track surface, or rather, how it affects the handling. I can see a dynamic, but global, grip level being fairly easy to implement. Pooling at certain points also fairly easy (though how these change with time is a bit harder).
Dynamic drying due to vehicles is one hell of a task to get right. It's the physical displacement of the water and the heating from the tyres own heat plus the friction on the surface (assuming the water is at the same temperature as the track - - gets ever-more complicated) that is the real challenge and, arguably, where the meat of the convincingness of the effect will be decided, so it needs to be right.

You'd essentially have to divide the surface of the track up into elements and solve each patch every frame for water incident from the rainfall, plus displacement from neighbouring elements (forced and under gravity) and likewise leaving itself. This gives an idea of "water-level" in the element.
Then you need to figure out its temperature based on the results from the first part, plus friction from any tyres, incident heat from the sun etc. etc.

So, we're talking about running FEM at the same time as playing the game. Not gonna happen.

CHEAT!

What can happen, though, is these simulation steps can be taken in pre-production and then parameterised according to the current "water-level" and the current rainfall rate. Then you'd have the equivalent of many "pre-baked" lightmaps, but for rain! You interpolate between these levels of rainyness for the modeling of the surface conditions according to the rainfall history only.
The bonus here is, depending on memory limitations, you could add in rubbering-in and "dust" effects really really easily, and couple it with temperatures.

Lots of testing will determine the resolution of the initial solver and the tables which get "looked-up" in-game for the "grip modifiers" of the track surface elements, according to temperature, rain and rubbered-in-ness. Well that's how I'd do it, anyway.

But, it all seems as though it's probably a bit much for GT.
And I wouldn't be satisfied with a half-arsed dynamic implementation, so as far as I'm concerned weather should either be fully dynamic or static.

Wow, I hurt myself.

 

[Stonemonkey]

VISUAL ELEMENT

I'd totally forgotten about the wet tracks in GT3 and GT4.

The SSR5 wet in GT3 was time-trial only, if I remember correctly. Anyway, it looked superb, but there was no precipitation, and I assume (can't remember) the grip level was constant over the entire track.

In GT4 we got a series of "special conditions" races at Tsukuba where you raced against one opponent. Again no precipitation, and I don't recall a difference in grip, although I may have been convinced of it at the time (this was probably due to my inconsistent driving style and eagerness to blame the game...)

So, the visual element has been working for some time, but has obviously incurred quite a performance penalty on the old hardware. The "shaders" etc. will need recompiling / re-optimising / re-writing for the new hardware, but it's essentially done.
What's missing is precipitation, something that is very hard to do convincingly without faking it with a lot of fogging (wait, that's all it is, right? Heavy, falling fog )

PS2 doesn't have shaders, well not pixel shaders at least. Wet surfaces need some sort of reflection so it's probably some sort of environment mapping applied to the track which will be done much more convincingly with shaders on the PS3. You're right about precipitation, a lot of people go on about how other games have done it so it should be easy but while you recognise it for what it's meant to represent it often doesn't look all that great.

PHYSICAL ELEMENT (a hypothetical rambling...)

Then of course, the modeling of the physical aspect on the track surface, or rather, how it affects the handling. I can see a dynamic, but global, grip level being fairly easy to implement. Pooling at certain points also fairly easy (though how these change with time is a bit harder).
Dynamic drying due to vehicles is one hell of a task to get right. It's the physical displacement of the water and the heating from the tyres own heat plus the friction on the surface (assuming the water is at the same temperature as the track - - gets ever-more complicated) that is the real challenge and, arguably, where the meat of the convincingness of the effect will be decided, so it needs to be right.

You'd essentially have to divide the surface of the track up into elements and solve each patch every frame for water incident from the rainfall, plus displacement from neighbouring elements (forced and under gravity) and likewise leaving itself. This gives an idea of "water-level" in the element.
Then you need to figure out its temperature based on the results from the first part, plus friction from any tyres, incident heat from the sun etc. etc.

I've kind of thought about this before, in the TT demo the tyre marks on the track look to me like marks that have been drawn with anti-aliasing onto lo-res maps and then applied to the track with a lot of filtering, if it's possible to stream them to/from the HDD or compress/uncompress in memory as you drive around onto different sections of the track it might be possible to have dynamic rubber maps, not only rubber but like you have red/green/blue in textures you could also maybe have dynamic wet maps. It'd be great to think so but whether something like that is possible on the PS3 with everything else going on in the game or not, I couldn't say.

So, we're talking about running FEM at the same time as playing the game. Not gonna happen.

CHEAT!

What can happen, though, is these simulation steps can be taken in pre-production and then parameterised according to the current "water-level" and the current rainfall rate. Then you'd have the equivalent of many "pre-baked" lightmaps, but for rain! You interpolate between these levels of rainyness for the modeling of the surface conditions according to the rainfall history only.
The bonus here is, depending on memory limitations, you could add in rubbering-in and "dust" effects really really easily, and couple it with temperatures.

Lots of testing will determine the resolution of the initial solver and the tables which get "looked-up" in-game for the "grip modifiers" of the track surface elements, according to temperature, rain and rubbered-in-ness. Well that's how I'd do it, anyway.

But, it all seems as though it's probably a bit much for GT.
And I wouldn't be satisfied with a half-arsed dynamic implementation, so as far as I'm concerned weather should either be fully dynamic or static.

Wow, I hurt myself.

Lol, I hadn't read your last bit when writing the other bit.

 

[analog]

There was a significant difference in grip on wet Tsukuba in GT4 i recall, quite slippery and quite fun and sometimes very treacherous especially when braking.

 

[Griffith500]

PS2 doesn't have shaders, well not pixel shaders at least. Wet surfaces need some sort of reflection so it's probably some sort of environment mapping applied to the track which will be done much more convincingly with shaders on the PS3. You're right about precipitation, a lot of people go on about how other games have done it so it should be easy but while you recognise it for what it's meant to represent it often doesn't look all that great.

Hence the " quotes! I'm pretty sure even the original Playstation was capable of delivering per pixel effects? The Playstation graphics hardware was, until the third iteration, a breed unto itself. But yeah, I've not ever seen a decent representation of rain in a game.

I've kind of thought about this before, in the TT demo the tyre marks on the track look to me like marks that have been drawn with anti-aliasing onto lo-res maps and then applied to the track with a lot of filtering, if it's possible to stream them to/from the HDD or compress/uncompress in memory as you drive around onto different sections of the track it might be possible to have dynamic rubber maps, not only rubber but like you have red/green/blue in textures you could also maybe have dynamic wet maps. It'd be great to think so but whether something like that is possible on the PS3 with everything else going on in the game or not, I couldn't say.

I recall the rubber marks looking "odd", but I never really gave it much thought, nor a proper look. If that really is the case, then the "texture" / lookup method may be a winner? Considering main memory limitations and its mediocre bandwidth, I would expect (no real basis for this assumption) that streaming is out of the question. So maybe a single map in memory with the SPEs doing a bit of on-the-fly "fudge" calculations might work? It's all a matter of testing, figuring out which bits can be faked and which bits must be properly simulated at run-time.

Lol, I hadn't read your last bit when writing the other bit.

So we're either both on the right track, or both insane!

@analog I meant that the grip didn't vary across the width, or along the length of the track - that it was one "value" for the entire track surface. There was, of course, a huge difference in grip relative to the dry version!

 

[hardvibes]

source?

Yamauchi-san

 

[Griffith500]

GT5P had 200k polygon cars while GT5 cars will have 500k.

Source?

Yamauchi-san

Not gonna end well. How about a link?

 

[Stonemonkey]

I'm pretty sure even the original Playstation was capable of delivering per pixel effects?

I think it's graphics hw was limited to affine mapping with gouraud shading and all without a depth buffer, I think anything per pixel would have to be done a pixel at a time with the cpu (33MHz?) which could probably only access video mem through the gpu. The PS2 was quite a bit better, perspective correction/depth buffer/multi texturing/vertex shading/bi-linear filtering but still no pixel shading afaik.

I recall the rubber marks looking "odd"

There were times when I thought I could make out quite large jaggies, but very blurred.

 

[hardvibes]

Not gonna end well. How about a link?

Kazunori Yamauchi: “We had maybe 300 polygons in a car in the first game. Now we have about 500,000 polygons in each car. Back then, pieces of the car were more like symbols. Now they are real and reflect light.”

http://www.examiner.com/examiner/x-12218-Video-Game-Examiner~y2010m1d14-GT5-500k-polygons-in-a-car
http://games.venturebeat.com/2010/0...-takes-a-fifth-stab-at-a-perfect-racing-game/
http://www.gameguru.in/driving/2010...-80-percent-of-ps3-processing-power-yamauchi/
http://www.vgchartz.com/forum/thread.php?id=98090
and so on ...



and about the 200.000 in GT5P

Whereas cars in Gran Turismo 4 were comprised of approximately 4,000 polygons, cars in the fifth game in the series have been bumped up to an astonishing 200,000 polygons.

http://uk.videogames.games.yahoo.com/ps3/previews/gran-turismo-5-prologue-f1f672.html

Kikizo: Certainly, this demo you have shown us looks phenomenal, please can you tell us about some of the background to achieving this standard on PS3?
Yamauchi: On average, the car models in GT5 Prologue is 200,000 polygons - this is fifty times more than what we used to have on GT4 PS2, which was on average 4,000 - so fifty times upscaled.

http://archive.videogamesdaily.com/features/gt5_kazyamauchi_iv_oct07_p1.asp

 

[Stonemonkey]

I find this a little confusing:

Kazunori Yamauchi: “We had maybe 300 polygons in a car in the first game. Now we have about 500,000 polygons in each car. Back then, pieces of the car were more like symbols. Now they are real and reflect light."

The only thing I can think of what he's trying to explain is that originally the simple 300 poly car model just had a texture wrapped around it and all materials that the car was made of were just part of that texture and they didn't have any individual properties, whereas now all the parts of the car that are made of different materials are seperate parts of the model and have their own material properties.

 

[speedfreak69]

I find this a little confusing:



The only thing I can think of what he's trying to explain is that originally the simple 300 poly car model just had a texture wrapped around it and all materials that the car was made of were just part of that texture and they didn't have any individual properties, whereas now all the parts of the car that are made of different materials are seperate parts of the model and have their own material properties.

Precisely.

 

[Quency]

PHYSICAL ELEMENT (a hypothetical rambling...)

Then of course, the modeling of the physical aspect on the track surface, or rather, how it affects the handling. I can see a dynamic, but global, grip level being fairly easy to implement. Pooling at certain points also fairly easy (though how these change with time is a bit harder).
Dynamic drying due to vehicles is one hell of a task to get right. It's the physical displacement of the water and the heating from the tyres own heat plus the friction on the surface (assuming the water is at the same temperature as the track - - gets ever-more complicated) that is the real challenge and, arguably, where the meat of the convincingness of the effect will be decided, so it needs to be right.

You'd essentially have to divide the surface of the track up into elements and solve each patch every frame for water incident from the rainfall, plus displacement from neighbouring elements (forced and under gravity) and likewise leaving itself. This gives an idea of "water-level" in the element.
Then you need to figure out its temperature based on the results from the first part, plus friction from any tyres, incident heat from the sun etc. etc.

So, we're talking about running FEM at the same time as playing the game. Not gonna happen.


Wow, I hurt myself.

Of course if you want to push it to the limits you must use FEM same as for damages, and being working in automotive engineering, to simulate a crash with a basic shell model (means using 2D plans instead of 3D solids) for a 200,000 elements model (equivalent to polygons) it takes nearly 10h of CPU calculation to simulate 1second crash lenght at 60 km/h...
If you want realistic crash modelisation in GT plug your PS3 to the NASA supercalculator

On the other hand we talking about race, which means that drivers are supposed to follow a racing "line" to be fast. Also a dynamic dry can be done by a simple algorithm on this racing line.

Let me explain myself:
If it rains, there is a coefficient which is the amount of water falling down per unit of area (can be cm²) if this coefficient is superior than the drying coefficient (amount of water which dries cause of tires or temperature per unit of area) then nothing happen, the road is wet and bad for addherence.

If the rain decreases then the track can start to dry, first on the racing line and then expand to the sides... that is probably what is used in V8 Superstar Next Challenge IMO and it wouldnt be too demanding in processing ressources !

 

[Kryz11]

And you base your guess on?

Since you talk about quality reduction, I was just talking what if PD reduces Frame rate? 20 FPS from 60 during RAIN wont really hurt the game.

 

[Stonemonkey]

Since you talk about quality reduction, I was just talking what if PD reduces Frame rate? 20 FPS from 60 during RAIN wont really hurt the game.

I'm not talking about quality reduction at all, as an extreme example, consider a car in the distance that maybe fills a total of 100 pixels on the sceen, is there any point in drawing that with 500,000 polys? Or would you use a simplified model? In fact, using a simplified model in that case could actually improve the quality and with some kind of dynamic LOD you could eliminate the popping.

Reduction from 60 to 20 fps would definitely hurt the game but I don't know what you're basing your numbers on anyway.

 

[Griffith500]

Of course if you want to push it to the limits you must use FEM same as for damages, and being working in automotive engineering, to simulate a crash with a basic shell model (means using 2D plans instead of 3D solids) for a 200,000 elements model (equivalent to polygons) it takes nearly 10h of CPU calculation to simulate 1second crash lenght at 60 km/h...
If you want realistic crash modelisation in GT plug your PS3 to the NASA supercalculator

On the other hand we talking about race, which means that drivers are supposed to follow a racing "line" to be fast. Also a dynamic dry can be done by a simple algorithm on this racing line.

Let me explain myself:
If it rains, there is a coefficient which is the amount of water falling down per unit of area (can be cm²) if this coefficient is superior than the drying coefficient (amount of water which dries cause of tires or temperature per unit of area) then nothing happen, the road is wet and bad for addherence.

If the rain decreases then the track can start to dry, first on the racing line and then expand to the sides... that is probably what is used in V8 Superstar Next Challenge IMO and it wouldnt be too demanding in processing ressources !

This would in theory work, but I'm not sure how acceptable it'd be - well it'd be fine assuming equal rainfall over the whole track, OK for Motorsports Land, but what about the Nürburgring? Sure, you could split the track up into sections, but in order to avoid nasty discontinuities / vague over-smoothed transitions, you'd be getting coverage pretty much to the point of the method I mentioned, just minus track-width considerations.

The other issue is one of racing line: what determines where it dries? Will it dry on the "AI-line" even though you use a completely different line (time trial, online...) ??

The main point of calculation in your method is analogous to mine: how to determine drying rate. I understand totally that from a performance standpoint, you want to be doing these calculations / lookups as infrequently as possible, but (as an engineer, I'm sure you understand) it's a matter of compromise against the fidelity / usefulness / "gameplay" addition.
Equally, we don't know without trying!

So to reiterate, I think we'll get a static weather implementation

As for crash damage, didn't Kaz say they'd played around with full deformation? I remember Carmageddon 2 had per-vertex deformation, and GRiD's is mightily similar, as is GTA IV's - only the latter two seem to have more sophisticated constraints. The main advantage in racing games is that chassis deformation can be heavily fudged (i.e. ignored for the most part, or triggered on a suitably large "event") and then only the body panels need be deformed (elastically at first) according to a simple "stiffness curve".

 

[Quency]

@Griffith500

I agree with you, whatever way you choose to render the wet track, how will it change and at what speed... And I know that even in real GT teams, it is really hard to determine !!! Knowing that they are all kind of engineers there, I doubt PD has enough knowledge and experience to do it.

For the driving line, although my suggestion was a really simple way to do it, although if you look at real races (I have in mind F1) even they dont have all the same racing line the track still dries on a small line first. But it comes to my mind that actually the braking lines were much dried and wide than the straight part...
This adds even more parameters to master !!!

Definitly your static weather implementation sound the most realistic.


As for damaged, I was talking about scientific (near to) exact deformation depending on material parameters... This is the "realistic" signification to me (bhooo stupid engineer !!! hehe)
Now I know that its possible to get some kind of really conveniant deformation quite easily and very fast (I remember spent a couple of weeks playing with 3ds Max crashing cars on a wall) just as water behavior in other games (which would recquires FEM in the same way)
It is as beautifull as inaccurate but at the end of the day for 99.5% of gamers it will look real (even for me actually).
So I have no worries that they will manage to do it very well and "sexy"


EDIT : Oh and just for information to those who find not realistic the behavior of the cars when hit the wall at 200 km/h, Crash test standards put the maximum vehicule speed for crash test to 63 km/h. Above this speed they are assuming that 90+% of population will be too severaly injured to "survive" !!! that's why they all concider F1's Kubica crash at Albert Park at Canadian GP three years ago as a miracle

- You should not crash at more than 63 km/h - hehe

 

[Stonemonkey]

Albert park's in Australia.

Kubicas crash in Canada was in excess of the requirements of the car but I think the requirements for F1 are a bit more than 63kph. Don't forget they have crash helmets and HANS and they're strapped into seats that are moulded to fit.

 

[Quency]

@Stonemonkey

sorry for albert park, and for F1 cars of course they have different requirements and the driver will actually die cause of the impact before the cockipt will get damaged. Human body cant stand more than 70G (means 70 times the gravity) on peak and less than 10G on constant. Kubica still had chance to survive to the impact.

Now if you look at "normal" cars, if drivers have hans and crash helmets its because they dont have airbag and because they are literally sealled to their seats. Also during an impact the force applied on their body is much higher than usual cause they follow the car spead/acceleration instead of having a smoother deceleration in case of normal belt+airbags.
Just to tell that whatever equipement you are using, above 60 km/h a crash involves so much of force applied to the body than it becomes lethal.

Edit: you must also see the driver's positon on the car. In F1 they are almost lying down instead of seating. Which improve the body's response on high acceleration (thats also why they can brake so hard, do it in a seating position and i can assure you will have some chest pain )

 

[Stonemonkey]

Just to tell that whatever equipement you are using, above 60 km/h a crash involves so much of force applied to the body than it becomes lethal.

If that was true then F1 teams would have a much higher turnover of drivers, they rarely drive slower than 60km/h and most have been involved in high speed crashes but I think the last driver killed was Senna 16 years ago.

 

[Quency]

If that was true then F1 teams would have a much higher turnover of drivers, they rarely drive slower than 60km/h and most have been involved in high speed crashes but I think the last driver killed was Senna 16 years ago.

Like I said before, Formula 1 is really different from the other cars. few things to consider during a F1 crash:
- First of all, the F1 shape itself make a frontal crash really difficult and tend to make the F1 to crash-slide on the wall.
- F1, like most of racing cars have an advance crash-box in the front part which is made to dissipate energy during crash (that's the main purpose of cars desintegration during crash).
- The driver is lying down which allows him to take more violent body stress.
- The drivers usually brakes (at least a bit) before crashing, so the car will definitly be at max speed (you can add the sand before they hit anything which is literally stopping the car)
- Drivers are physically more trained than any other driver which helps them to get more violent stress.
- Tracks are usually not made of walls, and tires barrier are acting as a big airbag and are placed on high risk sections.
-...

For so many reasons F1 and general race car are made in a way that allows this kind of crash to happen and not be too dangerous, but a big crash is almost always an entry ticket to the hospital for some time

Concerning normal cars like 60+% of GT5 cars, the 60 km/h is applied. After that you must consider each car differently. Like for an Integra then 60 km/h should be a correct lethal speed. For a Ferrari I think you can increase a bit the speed limit...

And I dont know if you already met with an accident but I can tell you that even a 30 km/h crash heavily damages your car, so if you want the game to be realistic, no need to go in a wall at high speed to terminate your race hehe

 

[cb1021]

I guess it's interesting that you mention that. If PD implements "realistic" damage, then you hit the wall at 60km/h, your car is going to get ripped apart... if you hit something straight on at even 30km/h, you're gonna see an airbag, no?

 

[Stonemonkey]

Just to tell that whatever equipement you are using, above 60 km/h a crash involves so much of force applied to the body than it becomes lethal.

Like I said before, Formula 1 is really different from the other cars. few things to consider during a F1 crash:
- First of all, the F1 shape itself make a frontal crash really difficult and tend to make the F1 to crash-slide on the wall.
- F1, like most of racing cars have an advance crash-box in the front part which is made to dissipate energy during crash (that's the main purpose of cars desintegration during crash).
- The driver is lying down which allows him to take more violent body stress.
.........

so why such a miracle?

EDIT : Oh and just for information to those who find not realistic the behavior of the cars when hit the wall at 200 km/h, Crash test standards put the maximum vehicule speed for crash test to 63 km/h. Above this speed they are assuming that 90+% of population will be too severaly injured to "survive" !!! that's why they all concider F1's Kubica crash at Albert Park at Canadian GP three years ago as a miracle

Whatever.

 

[Quency]

Whatever.

By "equipement" I was meaning Hans-Helmet/Belt-airbag... this was out of car consideration.
Each car has a different response to crash due to their conception, and like i said before 63km/h is a standard decided by crash-test association all around the world. It means it's the maximal speed at which a crash test is meaningful to set the car security towards crash, above this speed the result of crash is too hazardous and becomes (equals) lethal.

Sorry for the misunderstanding


EDIT: for Kubica's crash, they are considering as a miracle the fact that he end up the crash without really be injured cause of the violence of the crash when the car was litterally "flying" into the wall at 200+km/h. Massa did not had the same chance, and spent few weeks in the hospital (even if the most severe injure has been made by the spring hitting his helmet before the crash).