F12k That "floating" feeling.

A lot of people complain about the floating disconnected feel the F12k physics have, compared to GP3. They are right. The reason for this is because while the F12k physics engine is very sophisticated, the values fed to it are not realistic. And the reason is, I believe, is because the AI could not handle a realistic model. If you edit the values in the veh files for the cars the AI drive or alter the garage setups they use, it drastically changes their behaviour. This is a pretty cool and ambitious approach for AI, I don't know of any other sim that forces the AI to drive an actual car, but they just aren't sharp enough to handle realistic models. EA took the lazy and quick approach to deal with this problem. They chose a compromise. So the AI have a model that they can sort of handle, and you get a model that is sort of realistic. The right way to do it is to choose the car you want to drive. Then edit the veh file for it. Then edit all the AI veh files giving them an easier physics model. You have to choose the "not all cars identical" type options in the game menus. I got started on this approach after using some files and theories of I believe it was Gregor Veble? If you look in the veh files, there is a variable: inertia=(2000.00,2500.00,1000.00). If you think about it, why did ISI use a constant for inertia when there is no such thing? Why didn't they just let the mass and acceleration dictate what the inertia would be? It makes it easy to make tweaks that the AI can handle, and lets the suits dumb down the game for the mass market. The values represent pitch, yaw, and roll, around the center of the car, much like an airplane. This is why some people think the car rotates on its center axis when it turns. If you increase the yaw to 6000 it indeed does wobble back and forth around its center. Veble reduced them to what he thought a real F1 driver would feel: inertia=(900.00,1100.00,450.00). I believe that a value of 1000.0 gives the physics code a neutral multiplier of 1.0. So I use inertia=(1000.00,1000.00,1000.00). Now take a look at: engine_inertia=(0.190). This slows down the rev up speed when in neutral, and makes it easier for the AI (and you) to avoid spinning out. I use engine_inertia=(0.12) for faster revving. clutch_eff=(0.995). Dumbed down makes for smoother shifting and weaker transfer of power to the wheels. 1.2 will jolt the drivers head back when shifting and is wicked and fun. No way the AI can handle it. I use 1.1. Careful in the pits! >:) A high 1.2 value makes it hard to avoid spinning when hitting the curbs, but it also gives you a faster lap time. Now look at the inertia variables for the wheels. There a four of them, one for each wheel. Set to inertia=(2.000). This is not the rolling inertia. It is the SLIDING inertia. This greatly affects the abiblity to catch a slide like in GPL. The AI like higher values. inertia=(1.0) is probably neutral and the most realistic for modern F1. If you boost it up to 6 or even 12, and boost the pitch and roll up, while taking the yaw down to 1000, you get a feel very similar to GPL, especially with a more efficient clutch and hard tires. I use the values that I've stated are probably "neutral" and let the laws of physics dictate the results. I also like to reduce the tire stiffness down to about 37. This gives more grip. At first its hard to drive. The spins are sudden and "seem" somewhat canned. The understeer is more pronounced and you have to adjust your garage setup. You have to go with a slightly higher setup and be very alert when hitting the curbs. But after awhile it gets much more immersive and addictive. It's more frightening like in GPL. It seems more like what I see on TV. I can catch spins but not all of them, and when I do, I'm proud of it. I'll never go back. Tom

On 16 Aug 2000 19:32:33 GMT, [email hidden] (TToommLL) thus spoke:

Quoted text (15 lines)

>I use the values that I've stated are probably "neutral" >and let the laws of physics dictate the results. I >also like to reduce the tire stiffness down to about 37. >This gives more grip. At first its hard to drive. The >spins are sudden and "seem" somewhat canned. The >understeer is more pronounced and you have to adjust >your garage setup. You have to go with a slightly >higher setup and be very alert when hitting the curbs. >But after awhile it gets much more immersive and >addictive. It's more frightening like in GPL. It seems >more like what I see on TV. I can catch spins but >not all of them, and when I do, I'm proud of it. >I'll never go back. > >Tom

Interesting. Is it possible for you to zip up you're altered files so that I/others can try them out? I'm too lazy to do all that editing myself. :-) -- Nos

TToommLL <[email hidden]> wrote in message news:[email hidden]...

Quoted text (81 lines)

> A lot of people complain about the floating disconnected > feel the F12k physics have, compared to GP3. They are > right. The reason for this is because while the F12k > physics engine is very sophisticated, the values fed to it > are not realistic. And the reason is, I believe, is because the > AI could not handle a realistic model. If you edit the values > in the veh files for the cars the AI drive or alter the garage setups > they use, it drastically changes their behaviour. This is a > pretty cool and ambitious approach for AI, I don't know of any > other sim that forces the AI to drive an actual car, but they > just aren't sharp enough to handle realistic models. > > EA took the lazy and quick approach to deal with this problem. > They chose a compromise. So the AI have a model that they > can sort of handle, and you get a model that is sort of realistic. > > The right way to do it is to choose the car you want to drive. > Then edit the veh file for it. Then edit all the AI veh files giving > them an easier physics model. You have to choose the > "not all cars identical" type options in the game menus. > > I got started on this approach after using some files and theories > of I believe it was Gregor Veble? If you look in the veh files, there is > a variable: inertia=(2000.00,2500.00,1000.00). If you think about > it, why did ISI use a constant for inertia when there is no such thing? > Why didn't they just let the mass and acceleration dictate what the > inertia would be? It makes it easy to make tweaks that the AI can > handle, and lets the suits dumb down the game for the mass market. > The values represent pitch, yaw, and roll, around the center of the car, > much like an airplane. This is why some people think the car rotates > on its center axis when it turns. If you increase the yaw to 6000 it > indeed does wobble back and forth around its center. > Veble reduced them to what he thought a real F1 driver would feel: > inertia=(900.00,1100.00,450.00). > I believe that a value of 1000.0 gives the physics code a neutral > multiplier of 1.0. > So I use inertia=(1000.00,1000.00,1000.00). > > Now take a look at: engine_inertia=(0.190). > This slows down the rev up speed when in neutral, > and makes it easier for the AI (and you) to avoid > spinning out. I use engine_inertia=(0.12) > for faster revving. > > clutch_eff=(0.995). Dumbed down makes for smoother > shifting and weaker transfer of power to the wheels. > 1.2 will jolt the drivers head back when shifting and > is wicked and fun. No way the AI can handle it. > I use 1.1. Careful in the pits! >:) > A high 1.2 value makes it hard to avoid spinning when > hitting the curbs, but it also gives you a faster lap time. > > Now look at the inertia variables for the wheels. > There a four of them, one for each wheel. Set to > inertia=(2.000). This is not the rolling inertia. > It is the SLIDING inertia. This greatly affects the abiblity > to catch a slide like in GPL. The AI like higher > values. inertia=(1.0) is probably neutral and the > most realistic for modern F1. If you boost it up > to 6 or even 12, and boost the pitch and roll up, > while taking the yaw down to 1000, you get a feel > very similar to GPL, especially with a more > efficient clutch and hard tires. > > I use the values that I've stated are probably "neutral" > and let the laws of physics dictate the results. I > also like to reduce the tire stiffness down to about 37. > This gives more grip. At first its hard to drive. The > spins are sudden and "seem" somewhat canned. The > understeer is more pronounced and you have to adjust > your garage setup. You have to go with a slightly > higher setup and be very alert when hitting the curbs. > But after awhile it gets much more immersive and > addictive. It's more frightening like in GPL. It seems > more like what I see on TV. I can catch spins but > not all of them, and when I do, I'm proud of it. > I'll never go back. > > Tom > >

On the topic of the AI driving the actual cars. I don't think GPL makes the AI drive the set-up of a car as they can somehow swivel the car on the spot and stop from 5th in half the time it takes me and all that but i know GP2 and GP3 for that matter has the AI set-up built into the track. You can mess about with it in the track editor but I haven't really tried it but I'd presume that for instance giving them less rear wing will make them spin more, like in F12k.

Let me be the first to offer up a If it's in Nascar Heat, why isn't in all games" pots that we've becomes so accustomed to in reference to GPL... :) What if ISI\EA would have included different difficulty levels like MGI did? If the AI can't handle it, then make it only an option in multiplayer and pre-season testing (there's a ICR 1 term...) and say "THIS MODE IS TOO HARDCORE FOR ANYONE BUT EXPERT DRIVERS!!!" turning the failings of their AI into a feature. I suppose if us reasonable people made games, they'd be perfect... :) hehe, tis the advantage of looking at a product at end result, instead of how the developers have to look at it, kind of like when you write a paper, no matter how many times you read it, you miss a couple simple grammatical errors. TToommLL <[email hidden]> wrote in message news:[email hidden]...

Quoted text (88 lines)

> A lot of people complain about the floating disconnected > feel the F12k physics have, compared to GP3. They are > right. The reason for this is because while the F12k > physics engine is very sophisticated, the values fed to it > are not realistic. And the reason is, I believe, is because the > AI could not handle a realistic model. If you edit the values > in the veh files for the cars the AI drive or alter the garage setups > they use, it drastically changes their behaviour. This is a > pretty cool and ambitious approach for AI, I don't know of any > other sim that forces the AI to drive an actual car, but they > just aren't sharp enough to handle realistic models. > > EA took the lazy and quick approach to deal with this problem. > They chose a compromise. So the AI have a model that they > can sort of handle, and you get a model that is sort of realistic. > > The right way to do it is to choose the car you want to drive. > Then edit the veh file for it. Then edit all the AI veh files giving > them an easier physics model. You have to choose the > "not all cars identical" type options in the game menus. > > I got started on this approach after using some files and theories > of I believe it was Gregor Veble? If you look in the veh files, there is > a variable: inertia=(2000.00,2500.00,1000.00). If you think about > it, why did ISI use a constant for inertia when there is no such thing? > Why didn't they just let the mass and acceleration dictate what the > inertia would be? It makes it easy to make tweaks that the AI can > handle, and lets the suits dumb down the game for the mass market. > The values represent pitch, yaw, and roll, around the center of the car, > much like an airplane. This is why some people think the car rotates > on its center axis when it turns. If you increase the yaw to 6000 it > indeed does wobble back and forth around its center. > Veble reduced them to what he thought a real F1 driver would feel: > inertia=(900.00,1100.00,450.00). > I believe that a value of 1000.0 gives the physics code a neutral > multiplier of 1.0. > So I use inertia=(1000.00,1000.00,1000.00). > > Now take a look at: engine_inertia=(0.190). > This slows down the rev up speed when in neutral, > and makes it easier for the AI (and you) to avoid > spinning out. I use engine_inertia=(0.12) > for faster revving. > > clutch_eff=(0.995). Dumbed down makes for smoother > shifting and weaker transfer of power to the wheels. > 1.2 will jolt the drivers head back when shifting and > is wicked and fun. No way the AI can handle it. > I use 1.1. Careful in the pits! >:) > A high 1.2 value makes it hard to avoid spinning when > hitting the curbs, but it also gives you a faster lap time. > > Now look at the inertia variables for the wheels. > There a four of them, one for each wheel. Set to > inertia=(2.000). This is not the rolling inertia. > It is the SLIDING inertia. This greatly affects the abiblity > to catch a slide like in GPL. The AI like higher > values. inertia=(1.0) is probably neutral and the > most realistic for modern F1. If you boost it up > to 6 or even 12, and boost the pitch and roll up, > while taking the yaw down to 1000, you get a feel > very similar to GPL, especially with a more > efficient clutch and hard tires. > > I use the values that I've stated are probably "neutral" > and let the laws of physics dictate the results. I > also like to reduce the tire stiffness down to about 37. > This gives more grip. At first its hard to drive. The > spins are sudden and "seem" somewhat canned. The > understeer is more pronounced and you have to adjust > your garage setup. You have to go with a slightly > higher setup and be very alert when hitting the curbs. > But after awhile it gets much more immersive and > addictive. It's more frightening like in GPL. It seems > more like what I see on TV. I can catch spins but > not all of them, and when I do, I'm proud of it. > I'll never go back. > > Tom > > > > > > > > >

No gpl doesn't, The gpl AI drive within set lines on the track towards set base lines and set advised speeds at 1.0000 of their speed, they can go faster or slower depending only on the parameters within gpl_ai.ini and driver.ini Imagine dragging a toy car on a piece of string, it is applied with equal physics to real life, but it is being yanked to a set point, and does not need to be pointing the right way to get pulled, 'it' has no say over when pulled, only the recovery time set in the ini files does. -- Tim, [email hidden]|[email hidden] http://legends.sports-gaming.com/forums.shtml FREE forums for your league... http://legends.sports-gaming.com Sim racing website loosely based on gpl...

HEY TOM Once again you personnally raised the game for everyone KEEP IT UP wonder what does the top brass at EA sports really think about the game ?? or games in general, I mean are they interested at all .................... TToommLL <[email hidden]> wrote in message news:[email hidden]...

Quoted text (88 lines)

> A lot of people complain about the floating disconnected > feel the F12k physics have, compared to GP3. They are > right. The reason for this is because while the F12k > physics engine is very sophisticated, the values fed to it > are not realistic. And the reason is, I believe, is because the > AI could not handle a realistic model. If you edit the values > in the veh files for the cars the AI drive or alter the garage setups > they use, it drastically changes their behaviour. This is a > pretty cool and ambitious approach for AI, I don't know of any > other sim that forces the AI to drive an actual car, but they > just aren't sharp enough to handle realistic models. > > EA took the lazy and quick approach to deal with this problem. > They chose a compromise. So the AI have a model that they > can sort of handle, and you get a model that is sort of realistic. > > The right way to do it is to choose the car you want to drive. > Then edit the veh file for it. Then edit all the AI veh files giving > them an easier physics model. You have to choose the > "not all cars identical" type options in the game menus. > > I got started on this approach after using some files and theories > of I believe it was Gregor Veble? If you look in the veh files, there is > a variable: inertia=(2000.00,2500.00,1000.00). If you think about > it, why did ISI use a constant for inertia when there is no such thing? > Why didn't they just let the mass and acceleration dictate what the > inertia would be? It makes it easy to make tweaks that the AI can > handle, and lets the suits dumb down the game for the mass market. > The values represent pitch, yaw, and roll, around the center of the car, > much like an airplane. This is why some people think the car rotates > on its center axis when it turns. If you increase the yaw to 6000 it > indeed does wobble back and forth around its center. > Veble reduced them to what he thought a real F1 driver would feel: > inertia=(900.00,1100.00,450.00). > I believe that a value of 1000.0 gives the physics code a neutral > multiplier of 1.0. > So I use inertia=(1000.00,1000.00,1000.00). > > Now take a look at: engine_inertia=(0.190). > This slows down the rev up speed when in neutral, > and makes it easier for the AI (and you) to avoid > spinning out. I use engine_inertia=(0.12) > for faster revving. > > clutch_eff=(0.995). Dumbed down makes for smoother > shifting and weaker transfer of power to the wheels. > 1.2 will jolt the drivers head back when shifting and > is wicked and fun. No way the AI can handle it. > I use 1.1. Careful in the pits! >:) > A high 1.2 value makes it hard to avoid spinning when > hitting the curbs, but it also gives you a faster lap time. > > Now look at the inertia variables for the wheels. > There a four of them, one for each wheel. Set to > inertia=(2.000). This is not the rolling inertia. > It is the SLIDING inertia. This greatly affects the abiblity > to catch a slide like in GPL. The AI like higher > values. inertia=(1.0) is probably neutral and the > most realistic for modern F1. If you boost it up > to 6 or even 12, and boost the pitch and roll up, > while taking the yaw down to 1000, you get a feel > very similar to GPL, especially with a more > efficient clutch and hard tires. > > I use the values that I've stated are probably "neutral" > and let the laws of physics dictate the results. I > also like to reduce the tire stiffness down to about 37. > This gives more grip. At first its hard to drive. The > spins are sudden and "seem" somewhat canned. The > understeer is more pronounced and you have to adjust > your garage setup. You have to go with a slightly > higher setup and be very alert when hitting the curbs. > But after awhile it gets much more immersive and > addictive. It's more frightening like in GPL. It seems > more like what I see on TV. I can catch spins but > not all of them, and when I do, I'm proud of it. > I'll never go back. > > Tom > > > > > > > > >

... >TToommLL <[email hidden]> wrote in message >news:[email hidden]...... >> I got started on this approach after using some files and theories >> of I believe it was Gregor Veble? If you look in the veh files, there is >> a variable: inertia=(2000.00,2500.00,1000.00). If you think about >> it, why did ISI use a constant for inertia when there is no such thing? >> Why didn't they just let the mass and acceleration dictate what the >> inertia would be? Actually, intertia is constant, isn't it, if the body is not deforming? Even so, for movement, mass IS intertia. F=m.a, and F=I.a ... >> Now take a look at: engine_inertia=(0.190). >> This slows down the rev up speed when in neutral, >> and makes it easier for the AI (and you) to avoid >> spinning out. I use engine_inertia=(0.12) >> for faster revving. This intertia is probably the drivetrain's intertia (not wanting to rotate). Does 0.12 feel better? (uhm, when comparing a bit to what you see in real live on the tv screen?) ...

Quoted text (7 lines)

>> Now look at the inertia variables for the wheels. >> There a four of them, one for each wheel. Set to >> inertia=(2.000). This is not the rolling inertia. >> It is the SLIDING inertia. This greatly affects the abiblity >> to catch a slide like in GPL. The AI like higher >> values. inertia=(1.0) is probably neutral and the >> most realistic for modern F1.

Inertia for a wheel with mass at the edge of the wheel would be r^2*m (r=radius of wheel, m=mass of wheel). So for r=0.5m (bit large, ok) and m=20 (kg) you get I=0.25*20=5. Intertia can be calculated. Strange. F1-2000 .veh files seem to have a lot of nice physics variables. ... >> I use the values that I've stated are probably "neutral" >> and let the laws of physics dictate the results. Neutral isn't always 1.0. Especially not in real life. I'm not suggesting your changes are bad; that depends a bit on how the F1-2K physics engine handles those numbers. But some variables are just empirically deduced, and there would be no normal way of doing it in software (tire grip for example is much too complicated; you'll have to approximate). But if it gives a better feeling of driving, then all the better. Ruud van Gaal, GPL Rank +53.53 MarketGraph / MachTech: http://www.marketgraph.nl Pencil art: http://www.marketgraph.nl/gallery

TToommLL wrote:

Quoted text (18 lines)

> > I got started on this approach after using some files and theories > of I believe it was Gregor Veble? If you look in the veh files, there is > a variable: inertia=(2000.00,2500.00,1000.00). If you think about > it, why did ISI use a constant for inertia when there is no such thing? > Why didn't they just let the mass and acceleration dictate what the > inertia would be? It makes it easy to make tweaks that the AI can > handle, and lets the suits dumb down the game for the mass market. > The values represent pitch, yaw, and roll, around the center of the car, > much like an airplane. This is why some people think the car rotates > on its center axis when it turns. If you increase the yaw to 6000 it > indeed does wobble back and forth around its center. > Veble reduced them to what he thought a real F1 driver would feel: > inertia=(900.00,1100.00,450.00). > I believe that a value of 1000.0 gives the physics code a neutral > multiplier of 1.0. > So I use inertia=(1000.00,1000.00,1000.00). >

Actually, I think it was a simultaneous approach of many. There is a comment I would like to make to your post, though. These inertia values are, as you mentioned, the values of inertia for rotation around the three axes. These values are indeed constant, and are directly related to the distribution of mass around the center of gravity but not the acceleration of the car. (physics note: the inertia of a rigid body is not a simple number such as mass, but is actually a tensor that can be represented by a 3x3 matrix. The three values in the .veh file are the values of this tensor along the three main body axes, which are probably presumed to be the eigenaxes of the tensor (a slight simplification of the part of physics engine) ). You can estimate the inertia values by the dimensions of the cars and their mass, which is what I did. The results suggest that the values used are not the multipliers, but the actual values of the tensor in standard metric units. Furthermore, the roll inertia (the third) should be even lower according to my estimate, but set it to 300 or less and watch the wiggle of the AI cars ;) (they seem to have less timestep accuracy in their physics calculations). -Gregor

ack, now I remember why I failed Physics 210 in college. reading this gives me a headache, hehe. "Ruud van Gaal" <[email hidden]> wrote in message news:[email hidden]... > ... > >TToommLL <[email hidden]> wrote in message > >news:[email hidden]... > ... > >> I got started on this approach after using some files and theories > >> of I believe it was Gregor Veble? If you look in the veh files, thereis

Quoted text (52 lines)

> >> a variable: inertia=(2000.00,2500.00,1000.00). If you think about > >> it, why did ISI use a constant for inertia when there is no such thing? > >> Why didn't they just let the mass and acceleration dictate what the > >> inertia would be? > > Actually, intertia is constant, isn't it, if the body is not > deforming? Even so, for movement, mass IS intertia. > F=m.a, and F=I.a > > ... > >> Now take a look at: engine_inertia=(0.190). > >> This slows down the rev up speed when in neutral, > >> and makes it easier for the AI (and you) to avoid > >> spinning out. I use engine_inertia=(0.12) > >> for faster revving. > > This intertia is probably the drivetrain's intertia (not wanting to > rotate). Does 0.12 feel better? (uhm, when comparing a bit to what you > see in real live on the tv screen?) > > ... > >> Now look at the inertia variables for the wheels. > >> There a four of them, one for each wheel. Set to > >> inertia=(2.000). This is not the rolling inertia. > >> It is the SLIDING inertia. This greatly affects the abiblity > >> to catch a slide like in GPL. The AI like higher > >> values. inertia=(1.0) is probably neutral and the > >> most realistic for modern F1. > > Inertia for a wheel with mass at the edge of the wheel would be r^2*m > (r=radius of wheel, m=mass of wheel). > So for r=0.5m (bit large, ok) and m=20 (kg) you get I=0.25*20=5. > Intertia can be calculated. Strange. F1-2000 .veh files seem to have a > lot of nice physics variables. > > ... > >> I use the values that I've stated are probably "neutral" > >> and let the laws of physics dictate the results. > > Neutral isn't always 1.0. Especially not in real life. I'm not > suggesting your changes are bad; that depends a bit on how the F1-2K > physics engine handles those numbers. But some variables are just > empirically deduced, and there would be no normal way of doing it in > software (tire grip for example is much too complicated; you'll have > to approximate). > > But if it gives a better feeling of driving, then all the better. > > > Ruud van Gaal, GPL Rank +53.53 > MarketGraph / MachTech: http://www.marketgraph.nl > Pencil art: http://www.marketgraph.nl/gallery

yeah I tried out with lower inertia ( not intertia ) and found the car actually drives faster and has less push into corners Which all in all translate to lower mass. ie Now your cars has higher power to mass ratio which is actually a cheat. ALL said and done I still prefer the original That is my choice Ruud van Gaal <[email hidden]> wrote in message news:[email hidden]... > ... > >TToommLL <[email hidden]> wrote in message > >news:[email hidden]... > ... > >> I got started on this approach after using some files and theories > >> of I believe it was Gregor Veble? If you look in the veh files, thereis

Quoted text (52 lines)

> >> a variable: inertia=(2000.00,2500.00,1000.00). If you think about > >> it, why did ISI use a constant for inertia when there is no such thing? > >> Why didn't they just let the mass and acceleration dictate what the > >> inertia would be? > > Actually, intertia is constant, isn't it, if the body is not > deforming? Even so, for movement, mass IS intertia. > F=m.a, and F=I.a > > ... > >> Now take a look at: engine_inertia=(0.190). > >> This slows down the rev up speed when in neutral, > >> and makes it easier for the AI (and you) to avoid > >> spinning out. I use engine_inertia=(0.12) > >> for faster revving. > > This intertia is probably the drivetrain's intertia (not wanting to > rotate). Does 0.12 feel better? (uhm, when comparing a bit to what you > see in real live on the tv screen?) > > ... > >> Now look at the inertia variables for the wheels. > >> There a four of them, one for each wheel. Set to > >> inertia=(2.000). This is not the rolling inertia. > >> It is the SLIDING inertia. This greatly affects the abiblity > >> to catch a slide like in GPL. The AI like higher > >> values. inertia=(1.0) is probably neutral and the > >> most realistic for modern F1. > > Inertia for a wheel with mass at the edge of the wheel would be r^2*m > (r=radius of wheel, m=mass of wheel). > So for r=0.5m (bit large, ok) and m=20 (kg) you get I=0.25*20=5. > Intertia can be calculated. Strange. F1-2000 .veh files seem to have a > lot of nice physics variables. > > ... > >> I use the values that I've stated are probably "neutral" > >> and let the laws of physics dictate the results. > > Neutral isn't always 1.0. Especially not in real life. I'm not > suggesting your changes are bad; that depends a bit on how the F1-2K > physics engine handles those numbers. But some variables are just > empirically deduced, and there would be no normal way of doing it in > software (tire grip for example is much too complicated; you'll have > to approximate). > > But if it gives a better feeling of driving, then all the better. > > > Ruud van Gaal, GPL Rank +53.53 > MarketGraph / MachTech: http://www.marketgraph.nl > Pencil art: http://www.marketgraph.nl/gallery

Nos wrote; > >Interesting. Is it possible for you to zip up you're altered files so >that I/others can try them out? I'm too lazy to do all that editing >myself. :-) > Heh, I can understand that! I'm still fiddling with them myself. In fact, as a result of this thread, I've since made changes. When satisfied, I'll upload something. Tom

Ruud van Gaal wrote;

Quoted text (10 lines)

>>> Now take a look at: engine_inertia=(0.190). >>> This slows down the rev up speed when in neutral, >>> and makes it easier for the AI (and you) to avoid >>> spinning out. I use engine_inertia=(0.12) >>> for faster revving. > >This intertia is probably the drivetrain's intertia (not wanting to >rotate). Does 0.12 feel better? (uhm, when comparing a bit to what you >see in real live on the tv screen?) >

Yes. In fact at the start of an F1 2000 race, I can hear sound samples of engines revving up much faster than possible with 0.19. >Inertia for a wheel with mass at the edge of the wheel would be r^2*m >(r=radius of wheel, m=mass of wheel). >So for r=0.5m (bit large, ok) and m=20 (kg) you get I=0.25*20=5. >Intertia can be calculated. Strange. F1-2000 .veh files seem to have a >lot of nice physics variables. So I was wondering, if inertia can be calculated, why didn't they just calculate it, since they use mass and radius variables in the veh files. For the #11 Benetton radius=0.32 and mass =24.0 so if r^2*m then inertia=2.4576. So why inertia set to 2.0? > But some variables are just >empirically deduced, and there would be no normal way of doing it in >software (tire grip for example is much too complicated; you'll have >to approximate). Well, I've found that the radius, width and stiffness variables do effect grip and giving much wider tires not only gives more grip but makes it difficult to turn at the hairpins. So given that they model all that and the mass, what is the inertia variable needed for? Especially when the default EA inertia settings make the game feel like an arcade. Tom

On 22 Aug 2000 20:25:53 GMT, [email hidden] (TToommLL) thus spoke:

Quoted text (7 lines)

>Heh, I can understand that! I'm still fiddling with >them myself. In fact, as a result of this thread, >I've since made changes. When satisfied, I'll >upload something. > >Tom >

OK thx, looking forward to checking it out. -- Nos

Okay, so I have been educated. Thanks. So if the 3 inertia values describe the shape of the car, then what is the purpose of the inertia variables for the wheels? I find they greatly increases the slinky air-hockey track feel that I'm trying to get rid of, and set them down to 1. If AI behavior where not a consideration, how would you set the inertia values? It seems that the "floating" sensation is from a high inertia to mass ratio. Like a feather? So I tried again inertia=0.9,1.,0.45 but I increased the mass to 1000. This gives the car a more substantial on the road feel. Problem is, that in testing at Japan, I had to tighten the suspension to the max (in the garage, I haven't messed with the veh suspension yet.) Also, I doubled the aero downforce and drag coefficients, and increased tire grip so I could still handle the car thru the turns. It feels "solid". Like a machine on the ground, less like an airplane. Problem is, now the tire wear is very rapid, even with hard tires. I tried reducing tire friction, but the resulting sliding heats the tires up so fast, they are useless. But doubling tire mass seems to help, with initial testing. While the 'u' keypress still shows a high rate of wear, the tires remain effective and don't get so hot. I like the feel of this driving model, and it seems to have increased the FF a bit, getting pretty far off from the original though. Tom Gregor Veble wrote;

Quoted text (46 lines)

> >TToommLL wrote: >> >> I got started on this approach after using some files and theories >> of I believe it was Gregor Veble? If you look in the veh files, there is >> a variable: inertia=(2000.00,2500.00,1000.00). If you think about >> it, why did ISI use a constant for inertia when there is no such thing? >> Why didn't they just let the mass and acceleration dictate what the >> inertia would be? It makes it easy to make tweaks that the AI can >> handle, and lets the suits dumb down the game for the mass market. >> The values represent pitch, yaw, and roll, around the center of the car, >> much like an airplane. This is why some people think the car rotates >> on its center axis when it turns. If you increase the yaw to 6000 it >> indeed does wobble back and forth around its center. >> Veble reduced them to what he thought a real F1 driver would feel: >> inertia=(900.00,1100.00,450.00). >> I believe that a value of 1000.0 gives the physics code a neutral >> multiplier of 1.0. >> So I use inertia=(1000.00,1000.00,1000.00). >> > >Actually, I think it was a simultaneous approach of many. > >There is a comment I would like to make to your post, though. These >inertia values are, as you mentioned, the values of inertia for rotation >around the three axes. These values are indeed constant, and are >directly related to the distribution of mass around the center of >gravity but not the acceleration of the car. > >(physics note: the inertia of a rigid body is not a simple number such >as mass, but is actually a tensor that can be represented by a 3x3 >matrix. The three values in the .veh file are the values of this tensor >along the three main body axes, which are probably presumed to be the >eigenaxes of the tensor (a slight simplification of the part of physics >engine) ). > >You can estimate the inertia values by the dimensions of the cars and >their mass, which is what I did. The results suggest that the values >used are not the multipliers, but the actual values of the tensor in >standard metric units. Furthermore, the roll inertia (the third) should >be even lower according to my estimate, but set it to 300 or less and >watch the wiggle of the AI cars ;) (they seem to have less timestep >accuracy in their physics calculations). > >-Gregor >