The Big Forza 2 Thread *

The new Forza 2 gameplay video looks really good. Especially the lightning and the cars. I like that the tire traces stays on the track, makes a lot of diference;) .
 
I've said it before this thread and I'll say it again. Live For Speed is doing this kind of damage now. Sure it needs some tweaking, but its a good start.

That game doesn't use real car models, right? So that helps right there.
 
I don't believe those calculations need to be nearly that complex in order to achieve an acceptable result in a video game. Just like any other physical simulation you can greatly simplify the model at the expense of accuracy.

I fully expect a move away from pre-made damage models and one toward real-time deformation with the next generation of consoles.
Any sort of real time deformation, simplified or not, requires a way to calculate/estimate/guesstimate the distribution of forces through the gridpoint mesh. You can use an ugly geometry mesh instead of a neater FEM mesh, you can simply things out the wazoo (just make everything spring constrained instead of plastic), and can even cheat a bit by ignoring huge portions of the mesh and regenerating geometry on the fly to fill in (can that even be called a cheat!?!).

But that is still a huge, huge, huge, huge, huge computational problem to do at even a few frames per second. You're talking about geometry tesselation and geometry generation real time, based on some physical calculation for grid point deformations.

Erm, yeah. Not going to happen soon IMO, but I'd love to be proven wrong. Improvements by using a lot more pre-computed interpolation models and vastly more possible outcomes are more likely.
 
I've said it before this thread and I'll say it again. Live For Speed is doing this kind of damage now. Sure it needs some tweaking, but its a good start.

I'll have to go do some reading. I'm very skeptical that they are generating new geometry on the fly, but if so I'll be as amazed as anyone!
 
I'll have to go do some reading. I'm very skeptical that they are generating new geometry on the fly, but if so I'll be as amazed as anyone!

Its not great, the game's still in beta. But it car deforms relative to the hit it takes.

That game doesn't use real car models, right? So that helps right there.

And I'm not saying that forza should use it and going off my nut, I understand the licensing issues involved. I'm just pointing out that it can be done, and I'm sure will be done in an AAA soon. Once they find a good implementation of it.
 
did you guys read this?

there was no way these car manufacturers (with the current licensing) were going to approve random damage without having approval of each of the damage models in advance (pre-rendered) IMO.

as for the the new video...looks great.... imagine that, a game looking better and better closer to release than it did on exposure of the first screen shots. ;)
 
Its not great, the game's still in beta. But it car deforms relative to the hit it takes.
I read about as much as I could find on the subject, and I guess that's just no substitute for playing the game and making direct observations. It seems that the suspension is ususing FEM for damage modeling, which makes sense - a bar is a single element and calculations are fairly straightforward, having a dozen or so suspension bars to calculate length changes for in a crash wouldn't be too complutationally intensive.

Not very realistic, though, if that is what you are getting after. Bars don't change length in crashes, they deform plastically or break. The effects on steering and suspension in LFS are entirely artifical with no connection to how that same suspension deformed in reality would behave. In fact, that damage was put in the game only to stop people from intentionally wrecking or using cheats (wall riding, curb jumping, and such) that detract from the simulation experience (sound familiar?).

Body damage has no effect at all on gameplay - not aerodynamically or mechanically. It is visual only, and most people seem to complain about how wonky it looks (I'd like to play it myself to see... sounds like a pretty good sim all around). I'm not sure, since I can't find a screenshot of a damaged LFS car, but they may be using a few length changing bars for a simple chassis too with the body broken into pannels and tacked to those bars such that it follows the chassis motion. That would be anything but good or realistic looking, but certainly interesting. Do you have any screenshots so that I can take a look at how they are accomplishing this?

In short, I don't think LFS is a good standard to use as far as making damage "look" more realistic. If you want it to act realistic... maybe, but precomputed models with more intermediate damage stages is still by far the easiest way to get there for body damage.

The suspension modeling does sound interesting though. I think there is some merit, and with higher complexity suspension FEM's and more computational power you might get something that gave a decent approximation of real damage corresponding to force vectors. Definitely worth keeping an eye on, and hopefully LFS S3 will improve on this a lot.
 
At some point, you guys might want to consider getting a car and doing trackdays. You'll only achieve so much "realism" playing a video game :)
 
At some point, you guys might want to consider getting a car and doing trackdays. You'll only achieve so much "realism" playing a video game :)
I had my fill of that in FSAE. I like the thrill, but my sense of mortality is too high for the kind of racing that really entertains me! I think all my hair would turn white and/or fall out completely. I guess there's a difference between skill/talent (which I have enough of to not look stupid, and presume I could work on) and nerves (which seem to have withered away en masse).

:)
 
Any sort of real time deformation, simplified or not, requires a way to calculate/estimate/guesstimate the distribution of forces through the gridpoint mesh. You can use an ugly geometry mesh instead of a neater FEM mesh, you can simply things out the wazoo (just make everything spring constrained instead of plastic), and can even cheat a bit by ignoring huge portions of the mesh and regenerating geometry on the fly to fill in (can that even be called a cheat!?!).

But that is still a huge, huge, huge, huge, huge computational problem to do at even a few frames per second. You're talking about geometry tesselation and geometry generation real time, based on some physical calculation for grid point deformations.

Erm, yeah. Not going to happen soon IMO, but I'd love to be proven wrong. Improvements by using a lot more pre-computed interpolation models and vastly more possible outcomes are more likely.

What if you pre-computed some damage models and then did an interpolation between them in real-time? The pre-computed models would act as keyframes ( "keymodels" would be more accurate probably) and then the ammount of force would dictate to what degree you altered from the previous model towards the next more-damaged one. Then the model that is arrived at would be the new starting point for further calculations. Since it would be rare that you would actually hit the exact ammount of damage to hit one of the pre-computed models, you would see a lot more variability in the damage shown on the cars. And if you could actually show the model change (a la the morphing effect for 2d images) you could slow down a replay and watch the impact cause the model to change shape. This would be a huge improvement, without having to cope with having to do all the calculations of a real physics-based model.

How doable to you think this approach would be? Or is it similarly computationally intensive?
 
Any sort of real time deformation, simplified or not, requires a way to calculate/estimate/guesstimate the distribution of forces through the gridpoint mesh. You can use an ugly geometry mesh instead of a neater FEM mesh, you can simply things out the wazoo (just make everything spring constrained instead of plastic), and can even cheat a bit by ignoring huge portions of the mesh and regenerating geometry on the fly to fill in (can that even be called a cheat!?!).

But that is still a huge, huge, huge, huge, huge computational problem to do at even a few frames per second. You're talking about geometry tesselation and geometry generation real time, based on some physical calculation for grid point deformations.

Erm, yeah. Not going to happen soon IMO, but I'd love to be proven wrong. Improvements by using a lot more pre-computed interpolation models and vastly more possible outcomes are more likely.

Didn't Full Auto already do realtime FEM deformation? It's at the core of the game, since that game sprung from ATI's XBox 360 collision tech demo.
 
What if you pre-computed some damage models and then did an interpolation between them in real-time? The pre-computed models would act as keyframes ( "keymodels" would be more accurate probably) and then the ammount of force would dictate to what degree you altered from the previous model towards the next more-damaged one. Then the model that is arrived at would be the new starting point for further calculations. Since it would be rare that you would actually hit the exact ammount of damage to hit one of the pre-computed models, you would see a lot more variability in the damage shown on the cars. And if you could actually show the model change (a la the morphing effect for 2d images) you could slow down a replay and watch the impact cause the model to change shape. This would be a huge improvement, without having to cope with having to do all the calculations of a real physics-based model.

How doable to you think this approach would be? Or is it similarly computationally intensive?
I think that's probably the way we will see damage modeling (graphically) advance over the next several years. More intermediate models produced by tesselation (on both front end to save artists time in producing key frames, and on the back end as a skinning/tesselation/animation sort of thing) and higher variety of damage.
 
Well, now, that's got to be one of the more beautiful shots I've seen. Though I can't help but assume that is some sort of photo-mode-souped-up-camera-angle-PR-pull-out-all-stops-on-the-engine shot and something that I will never see in the game.

But still very beautiful!
 
Great Post!

You do that once, and then you think - wait, for the price of one track-day, I can buy all the consoles, wheels, three racing games and a nice LCD TV, and drive as much as I like, on many different tracks, with hundreds of different cars I can modify and set-up however I like, whenever I like, without having to worry too much about killing myself.
 
You do that once, and then you think - wait, for the price of one track-day, I can buy all the consoles, wheels, three racing games and a nice LCD TV, and drive as much as I like, on many different tracks, with hundreds of different cars I can modify and set-up however I like, whenever I like, without having to worry too much about killing myself.

That fun from one trackday is easily worth the price of the consoles. Oh and trackdays around here are $150-$200 max for the whole day. As for killing yourself at the track.....you'll have to go out of your way to do it. Chances are generally higher that you'll die in a public road incident than a track incident.

If anyone is seriously interested in going fast and better yet, has a decent sports car, I'd highly recommend they put away the kiddie crap and check out a trackday. It's worth every penny.
 
You do that once, and then you think - wait, for the price of one track-day, I can buy all the consoles, wheels, three racing games and a nice LCD TV, and drive as much as I like, on many different tracks, with hundreds of different cars I can modify and set-up however I like, whenever I like, without having to worry too much about killing myself.

LOL, that's like arguing porn on your computer is better than sex with a real women. There's just no comparison ;)
 
Back
Top