Jump to content Jump to content

V2 Physics Discussion


griev0r

Recommended Posts

Thankyou for differentiating between pre-baked and real-time simulation, i hope it'll help many here.

1) I read that you were making porno videos, if i hadn't already finished my coffee, it would be all over the screen :(

2) It's important to note here, that because a rally game has far fewer possibilities than a flight sim, it's much easier to enter the pre-calculated info into the game.  I don't mean to say this is easy by any stretch, just not like a flight sim will be :D 
Tweaked that now. =p
It's true that there are kind of less possibilities in a rally sim, but that's what helps something like DiRT Rally be reasonable quite a lot of the time, whereas a flightsim tends to be reasonable only a little bit of the time (which is why something like MSFS is absolutely hopeless at high performance aircraft, be they acrobatic Red Bull Racers or military fighter jets. They need a simulator more geared towards them from the ground up like DCS).

The point is that simulation really isn't all that easy and what we have at home involves a lot of fudging.
It's not even as simple as putting in accurate data = accurate results.
You often have to use what, on paper, looks like very bad data and go by the the feel instead.
There isn't a single racing sim out there that also has good crash simulation.
Probably the best crash simulation going right now is BeamNG.drive, but that's because it's built around soft-body physics and deformation first and foremost, rather than the driving.

In the end, it's all about compromise. The v2 physics are a compromise to create a better driving experience, but at the cost of how the cars behave when they aren't on the ground.
v2 physics were a re-work of the v1 physics which is part of the problem.
Truly solving the airborne problems while maintaining the way the cars behave on the ground unfortunately requires a more substantial re-working of the physics engine than what v1 to v2 was.
That's what we can take away from the DiRT Show.
Link to post
Share on other sites
I think in late 2016 or early 2017 we'll see a Dirt Rally 2 that's completely polished and has more locations, features and a new & better engine that's capable of doing things proper. I can't see them tweaking to much at this point on an engine that they've come out and said they don't like (sort of), when they could just make a new one for future titles that fits their needs and our wants better. 
Link to post
Share on other sites
teknoid85 said:
ok,  thats a controlled jump with a studied exit ramp.     try this on finland with a stock truck and see what happens
It was an allusion to the fact that the trajectory of the flight depends on the velocity and aerodynamics, but almost not depends on the car weight. There is no point to discussing the obvious things about that the stock truck will break at landing.

Link to post
Share on other sites
BrySkye said:

Oh, you have no idea how far off the mark you are. No flight sim games actually simulate all of that the way you think.

This be a tangent, but it's an area I actually do know some of the inner workings of.
You know how the Microsoft Flight Simulator "Flight Dynamics Engine" works?
Look up tables. Not real time simulation. Hard numbers in what is basically a spread sheet.
Quoting just part of your post to save space not to take it out of context. 

Very interesting discussion this. I'm big on flight sims and specifically Falcon 4. I'm sure you have heard of BMS since you're into this stuff. When the new F-16 flight model was being developed for the sim (quite a few years back), the lead FM guy had a copy of NASA Technical Paper 1538 in hand. It is titled "Simulator Study of Stall/Post-Stall Characteristics of a Fighter Airplane With Relaxed Longitudinal Static Stability". It is basically a case study on RSS FBW aircraft using the F-16 as the test bed, I'm sure you can find it online if you want. I had taken a look at it and I won't pretend to have understood any of it, but the lead FM developer is actually an aerospace engineer and is very serious about getting this stuff right. The first iteration of the "Advanced Flight Model" (AFM) was basically fudging (in a way) the simulator FM and flight control system (FLCS) to get the output to match the NASA TP. As development went on and the environment and physics evolved as well, it got to the point were the real FLCS logic was coded into the sim, and the physics engine produced the desired output. This is the "as close as" part. Real F-16 pilots have also confirmed that the sim behaves as expected, and that especially the stall/post-stall behavior (where obviously there was an abundance of data from the NASA TP) it is "perfect"/"exactly like the real jet". Now, as far as I know, this is all real time simulation of aero, engine, FLCS and equations of motion. I distinctly remember the point in development after which look up tables (which F4 also used) were a thing of the past at least as far as the F-16 was concerned. Even if it was (or is if I don't remember correctly) look up tables though, this falls in the "makes no difference" part. If for any given input the output is correct, the inner workings are irrelevant IMO. 

I'm aware of the limitations of MSFS but that is one very specific example and it may even have more to do with the engine being old and limited development time instead of modern computing power exclusively. The main limitation MSFS modders face is that you have to work within the engine's shortcomings with no possibility to change anything that is hardcoded. That was not an issue for BMS and I think is the major difference. I'll wager the DCS physics engine and at least ED's in-house aircraft are at least on par with BMS. 

I'd love to discuss this further but I think we're taking this way off topic. In any case I think you will agree that DR doesn't need to come close to this level to fix whatever odd airtime behavior there might be. Keep in mind that in BMS the sim also has to simulate a whole campaign brain, ground/air/naval units AI, the jet's avionics (extremely complex on their own), hydraulic/electrical/fuel systems, weapon logic and FM, etc, etc. Now I'm not saying that it is easy to implement proper aero in DR, I'm not even saying that they should make it a priority or that they should do it at all. I'm just saying that, from a computing power standpoint, it can be done. 

Apologies to everyone for the long off topic post. 
Link to post
Share on other sites
Thing about BMS and further dev of Falcon 4 as a whole is that it's a passion project that's been on-going for 15 years now, made possibly by the sims source code being leaked. There's been years between 4.32 and 4.33 alone.
Things like schedules, deadlines, budgets and peoples wages aren't really an issue.
So it's not really a comparable scenario.

You could kind of compare it to the Richard Burns Rally scenario, which in a fully modded state is pretty much unrecognisable from the original game released 10 years ago. Things have been substantially re-written after many many years of experimentation and testing.

With DCS, you're looking at each aircraft (ie, module) costing as much as a whole game and each one is in development for a considerable amount of time.

To clarify though, both BMS and DCS still use 'programmed' physics and flight models, rather than actual real-time calculations based on the aircraft as physical objects. The closest to proper real-time calculated physics is X-Plane using Blade Element Theory, where instead of programming the flight dynamics, they are calculated/simulated based on the actual shape of the model.
Results are... varied.
Link to post
Share on other sites
Development time was not the point of the discussion, not to mention in both cases a proper software development studio with set working hours and schedules would have accomplished many times more if they kept steadily developing the same game for 15 years (not to detract from the amazing work both communities have done). As for 'proper' real time calculations, look up tables mean that everything is preprogrammed in the physics, like for this airspeed and altitude there is this amount of drag. In BMS (and DCS I imagine) the aero characteristics of the jet (drag index in this example) are saved  as values somewhere and the physics engine calculates the drag and how it affects the jet from a multitude of things including drag index in real time when flying. That's what real time is, programming the causes and letting the physics produce the effects, as opposed to programming the effects which is how look up tables work. The only difference in XP is that the values are derived from the 3D model instead of a file, and that was primarily aimed at helping developers create aircraft easier, not more accurate calculations. In fact as you say results are varied because as I understand it, it relies on an absolutely perfect model. 
Link to post
Share on other sites
Basically, what you're talking about isn't what is defined as real-time simulation.
BMS and DCS do not meet the definition of real time simulation. It's very highly refined pre-baked simulation. That's just a fact I'm afraid.
Actual real time simulation of aerodynamics, (or just about anything for that matter) at speeds high enough to be viable for a even slightly accurate game/simulation, requires significantly more processing power than is available in home computing.
There are reasons why such simulators require something more akin to server rooms.

Real-time simulation doesn't automatically mean more realistic/detailed either.
When an F1 team develops a new aerodynamic part, they don't spend months programming the values of that new part in the simulator. 
They drop it in, hoping their physics and simulation itself is accurate enough to give proper results.
Sometimes they do, sometimes they don't. Part of why they test those new parts in free practice and don't always use them in the race itself.
Hence the comparison to X-Plane.

So full real time simulation of aerodynamics/physics isn't on the cards for DiRT Rally, or many games for that matter.

It's possible to do very good programmed physics which are efficient enough to be viable (though the number of factors calculated is still an issue. There's a reason why more advanced home flight simulators tend to be CPU bottlenecked), but they do require time and that's what has basically ran out in DiRT Rally, as all explained in the DiRT Show.
Link to post
Share on other sites
Took the GrpA sub round Germany a few times and i really like what version2 gives us.

Although you can't feel the car sliding and stepping out(need FFB update) you know it's happening.
I think the car round Germany is more believable than version1, where it was just planted no matter what you did.

I also think once they sort out the FFB for tarmac, it'll be really good and a major step forward.
Although taking it round Finland, i can't help but think there's some sort of traction andor stability control going on,
even though i always have them turned off.

Good job Codies on version2.

ps. i'm not disagreeing with folks when it comes to the strange behaviour(distance travelled etc.) going over jumps.
Link to post
Share on other sites
BrySkye said:
Basically, what you're talking about isn't what is defined as real-time simulation.
BMS and DCS do not meet the definition of real time simulation. It's very highly refined pre-baked simulation. That's just a fact I'm afraid.
Actual real time simulation of aerodynamics, (or just about anything for that matter) at speeds high enough to be viable for a even slightly accurate game/simulation, requires significantly more processing power than is available in home computing.
There are reasons why such simulators require something more akin to server rooms.

Real-time simulation doesn't automatically mean more realistic/detailed either.
When an F1 team develops a new aerodynamic part, they don't spend months programming the values of that new part in the simulator. 
They drop it in, hoping their physics and simulation itself is accurate enough to give proper results.
Sometimes they do, sometimes they don't. Part of why they test those new parts in free practice and don't always use them in the race itself.
Hence the comparison to X-Plane.

So full real time simulation of aerodynamics/physics isn't on the cards for DiRT Rally, or many games for that matter.

It's possible to do very good programmed physics which are efficient enough to be viable (though the number of factors calculated is still an issue. There's a reason why more advanced home flight simulators tend to be CPU bottlenecked), but they do require time and that's what has basically ran out in DiRT Rally, as all explained in the DiRT Show.
It might not meet your definition mate, but that's what real time is. Real time physics means the actual physics (as in, equations of motion) are calculated in real time, not the object's physical attributes, like weight. (Actually that's not exactly the definition either, but by the formal definition all flight and racing sim games are real time so let's leave that aside for now.) What do you mean F1 teams just drop the aero part in? They drop it in where? Inside the server case? Of course they program its physical attributes in the sim, they just don't program the expected behaviors, the sim does that for them. That's what real time means in this context, that's what BMS does and that's the opposite of what MSFS did. Whether those physical attributes are derived from a text file or a 3D object file has nothing to do with the physics calculations. The look up tables way means there are no physics calculations in parts or all of the sim, it just takes the physical attributes as input and looks up the proper value to output without having to actually calculate it. If the physical attributes, while correct, are outside a certain range the devs had in mind when designing the look up tables (for instance fighter jets in MSFS), the physics break. In real time physics calculations, as long as the physical model is correct and the physical attributes are input correctly, the output will be calculated correctly too. 

You seem to be missing a link between the physical attributes (the wing's shape) and the behavior (what the wing does inside the airstream). The latter is part of the physics calculations, the former is not, whether XP derives it from the model has no bearing on it. Or you could be confusing two different aspects of physics simulation. One is used by engineers to develop parts, materials and systems, running structural integrity tests and the like. Indeed those are impossible to run in real time on desktops, however real time simulations that simulate an aircraft's behavior from the pilot's perspective are perfectly doable and not just as barely passable approximations. I could give you the e-mail of the BMS FM dev and you could tell him without taking one look at his code that what he did isn't real time or that he didn't calculate all the factors. But I'm sure you wouldn't care to actually listen to his explanation either. 

Your assumptions about the power required by professional sims are also just that, assumptions. You assume that because they are pro sims they must need more processing power exclusively for the physics, and because they have servers they must have physics that can't run on a modern desktop. You assume that the CPU bottleneck is due to the physics without having the code to run profiling tests. You also assume that because MSFS couldn't do it it is impossible in general. Just look at the avionics, they require a big computer to run inside the aircraft but we can fully simulate them on our poor little desktop processors alongside the rest of the game and whatever other processes might be running at the same time. I can also only assume, but if I had to guess I could easily think of many reasons why they need huge server rooms. First of all most of them are on huge motion platforms. They also need to drive the aircraft's actual physical instruments (not your monitor display) and avionics which, by the way, need massive cooling outside the aircraft since they can't be cooled by engine bleed air. Then there are instructor panels, telemetry systems and so on. 
Link to post
Share on other sites
You assume that I assume.

Ultimately, we're talking about very different things and you don't seem to get that. 

When I'm talking about 'true real time simulation' of things like aerodnyamics and physics, I'm talking about computational fluid dynamics.
Which is about a very robust engine where everything is handled by the simulation side.
You program the engine as opposed to assigning attributes to the objects put in that engine.
Not only is CFD far from infallible (ask Virgin Racing when they initially thought it could replace wind tunnels), but to run at sufficient speeds for gameplay purposes, yes, it needs either pretty hefty hardware or to be incredibly rudimentary. 
Of course you don't need a server room just to handle these calculations. That's a lot of twisting what I've been saying.
(well, actually that depends on just how many factors are party of the simulation and how quickly you need them calculated, but that should be obvious).
But the limitations in processing power to achieve acceptable game-level performance are very well documented.
Say, by Intel
https://software.intel.com/en-us/articles/fluid-simulation-for-video-games-part-1

If you want to say that any calculations done in real time make it a real simulation, than even something as archaic as MSFS counts.
I was trying to emphasise the difference between an engine which calculates everything for itself in real time and one which makes use of look up tables in a pre-programmed manner.

We''re done here. I've tried to keep DiRT Rally as heavily involved in this as possible, but we're beyond that now.
Feel absolutely free to keep thinking that @KickUp was talking BS in the DiRT Show.
Link to post
Share on other sites
How come my glorious picture vanished from my post? Silly forum. :( Well, thanks Ozores, I guess...
Believe or not is a ...lon... weird story.

I just got curiosity about what that picture contain, so I inspected the element with the brouser, take the link and put it on another tab.And picture was there, but I noticed there was a swearword on filename, so forum converted to **** ... and broke the link.

I searched for another one, put on the quote and voilá.

CSI mode -> OFF
(you owe me a jug of honey mead)

Link to post
Share on other sites
chukonu said:
What do you guys think of the R4's when it comes to jumps? To me, especially the Impreza WRX seems almost perfect... Doesn't jump nearly as much as for example saaaaay the Peugeot 205!
I must say I do enjoy the slower, heavier R4 cars in Finland. The jumps and crests feels like they have more flow to them when not flying as much.
Link to post
Share on other sites
BrySkye said:
1. You still have to assign attributes to the object the engine uses, they don't magically appear in the engine. 
2. I twisted nothing you said, if I misunderstood you must have been talking about something else while still presenting it as a counter argument to the original discussion. You said it can't be done on home PCs because real sims need huge servers, now you say the servers are not there for the physics necessarily. Where were you going with that argument then exactly?
3. RE: that Intel article, that's not entirely relevant to what I was saying but ok, I still didn't see the point were the author said it was impossible though. On the contrary he wrote about how it can be done. 
4. With no disrespect to any of the guys at CM, I'm not saying he's talking BS but I won't take someone's word who says it can't be done, without actually having tried to do it at that level and knowing the challenges, at face value. 
5. You tried to keep DR involved? My point that all this isn't even necessary for DR you didn't even try to acknowledge. 

You can keep thinking that the real time sims that actually exist do not exist or are not actually real time or if they are, they do not conform to your definition of real time where apparently the requirement is to be a bespoke military grade simulator and/or simulate the entire universe. Yes, we are talking about different things I'm sure, I'm talking about usable and accurate (for the user - NOT the engineer/designer which is what F1 sims cater to and yes they need much higher fidelity) simulators and you are talking about simulating every single particle in the air which is obviously both impossible as you say, and completely irrelevant to flight and racing sims. I was trying to have a civilized and informative discussion but you seem intent only on proving that you know more than me. Which might very well be true, but since you've added exactly nothing of concrete value to this discussion except the preconceived notion that it's too hard therefore impossible or insisting that BMS uses look up tables when it doesn't, I guess we'll never know. Or at least I won't, I'm sure you think you do. 

Now we're done. 
Link to post
Share on other sites
bogani said:
chukonu said:
What do you guys think of the R4's when it comes to jumps? To me, especially the Impreza WRX seems almost perfect... Doesn't jump nearly as much as for example saaaaay the Peugeot 205!
I must say I do enjoy the slower, heavier R4 cars in Finland. The jumps and crests feels like they have more flow to them when not flying as much.
Precisely, though, the Mitsibushi seems to suffer much more airtime than the Subaru which seems top-heavier (more like it should).
Link to post
Share on other sites
Alwert said:
THIS IS……REAL SIMULATOR !!! 

ASSETTO CORSA- Rally Mod

As opposed to the fake simulator that dirt rally is I assume? Would you like some food troll? I can't find another explanation for this pointless description of an otherwise interesting video.
I keep wondering why people get offended so easily when the criticism isn't written with flowers and rainbows... Alwert didn't say DiRT Rally was a fake simulator, but anyways...

Let's get the criticism with flowers and rainbows and poneys and colours and ribbons then:

The tarmac in DiRT Rally is a little bit too smooth, and the little ondulation and oscilation it has has been softened even further with v2 physics.

Enough homo stuff (not judgin'). From day one of v2 physics I pointed how this new approach would jeopardize the feeling of challenge DR has. Before, if we pass with one wheel over the grass in tarmac it would mean almost 100% sure of a spin. It was really really really really really difficult to regain control of the car, and IF we did end up recovering, the feeling of reward and awesome driving was, well, awesome. Since v2 physics, grass is beeing treated as a curb, where you can cut as much as you can to gain time, with no handling penalty. This leans the handling model towards previsous versions of Colin McRae titles.

You had to really keep it on your toes. It was easy to upset the cars and spin up. It was challenging. Now, no mather what I do, I just can't seem to upset the car. It is just sooooo smooth. Sometimes it almost feels like I'm playing Colin McRae DiRT from 2007, without, obviously, that jerky moves the front wheels had back then.

I consider myself a picky, demanding, rigorous, exigent person.With the latest releases I could easily go with the tide and point "Heeey that is Finalnd in Autumn" or "4 cars that were already in DiRT 3?!". But I didn't. You know why? Because I believe that the proposition of DiRT Rally is different. The proposition is to regain respect among rally lovers and show some serious business to the sport.

I'm putting aside the fact that we have only 2 stages per country and only 2 true new rally cars since DiRT 3 (not considering RX and PP) because the main purpose is to have something immersive, pure rally style. I'm even letting it go (pun) the fact that Paul was the Chief Game Designer of DiRT 3 and we had more licensed cars back then (what happened from 2011 to today?). Anyways...

Sounds and graphics are amazing. Physics WERE challenging until v2. At least on tarmac. People can say that it's not because it's easier that it's more unrealistic. Well, maybe fock realism, maybe I just wan't something truly challenging...

But hey, that's me...


Link to post
Share on other sites
Alwert said:
THIS IS……REAL SIMULATOR !!! 

ASSETTO CORSA- Rally Mod

As opposed to the fake simulator that dirt rally is I assume? Would you like some food troll? I can't find another explanation for this pointless description of an otherwise interesting video.
I keep wondering why people get offended so easily when the criticism isn't written with flowers and rainbows... Alwert didn't say DiRT Rally was a fake simulator, but anyways...


That was implied when he pointed out that "This is a real simulator". I have no problem with criticisms or your comments, I'm just not sure how insinuating Dirt Rally is not a simulator and simply stating "should improve so much" is helpful to anyone. I'm aware there are improvements that could be made and that the simulation has flaws.

Imagine if I was reviewing a movie and said, "you couldn't even call this a film, it should improve in every way!". That is a boring conversation and pointless. Generic "this sucks" comments are a waste of  time for whoever is reading them and the time of the writer themselves.

So I wasn't offended, I don't really care if ppl don't like the game as it doesn't affect how much I like it, but I do wonder what some ppl get out of crapping on things with no intention of actually improving them.
Link to post
Share on other sites
I also didn't understand what one should see in the corsa-video. Nothing interesting or relative to topic...

But I found another one (and much more watchable) at that channel. Real Wales!

https://www.youtube.com/watch?v=6DaTH1DN_zc


P.S.: If someone thinks that the developers should strive for such difficulty, then he seriously mistaken. No one will make a game (finished AAA game!) for a small group of fans of extreme difficulty.


Link to post
Share on other sites

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...