Gran Turismo 6 hotlap viewer/analysis

[MadMatt]

Following a break on my major laptime simulator and taking advantage of Gran Turismo 6's new Motec feature, I decided to create my own data analysis/viewing tool.

Basically, GT6 now allows you to export a hot lap in Motec format. Great! For the video side of things, you still need a capture card tho. But so I was able to get a hot lap in Motec, and export the data given by GT6 in a Matlab format, and make a small app that allows me to watch my replay with real time telemetry (the little that GT6 gives you). This doesn't forbid you to create your own channels of course.

Here is after 2 days of work the look of my GUI (open in new window):



The information given by the game itself are:

- Lap time G M
- Car speed G M
- Engine speed G M
- Gear G M
- Steering angle M
- Tyre wear G
- Total distance covered by your car G
- Fuel level G
- Boost level G
- Throttle position G M
- Brake position G M
- Lateral acceleration G M
- Longitudinal acceleration M
- Vertical acceleration M
- Traction control G
- Wheels speeds M

G means info given only in the game while playing, M info that can be seen in Motec.

So this is all interesting, I can now see my lap with extra information that the GT6 GUI doesn't give me, such as quantitative accelerations (you only see a numberless gauge in GT6), but also values of brake, throttle, and steering application. I will improve my GUI to make it really easy to see and use, and probably release it for the public at some point. I just took this pretext to do some Matlab work with new features that I had never used before.

Anyway, we all know that GT6 seems to be a bit unrealistic when it comes to physics. It is ok for a game with large public, but for purists, it is not so good. But how "not good" or bad is it. I am not here to demolish the game, because I am a huge fan, I bought GT1 on the PS1 at day 1 with the new dualshock and I've owned every episode since. But there are things that are clearly not well made. Take a look at this example below:



The car is a Nissan R92CP, one of my favourite car of the game (and real life, being a group C fan):



In GT6 you can enhance its engine (like it needs to). So here are the new numbers:

- Power: 1153bhp
- Weight: 900Kg (balance 49% front)

Quite a monster! The graph you see has been taken from acceleration tests I did on the 30km oval. I took a section where we are absolutely flat, no bump, no crest, no steering input needed. I simply accelerate from 0 to whatever speed. What I wanted to understand is how the traction control worked (yes I enabled traction control for this car, level 3 out of 10, so pretty low).

To start with, we see that the throttle application (green) reduces as the speed goes up. I was full throttle on my paddle, and we clearly see here the TC in action. However it doesn't respond to any logic. In yellow you see the longitudinal acceleration which should remain constant at a max value corresponding to the grip available, until the engine is not powerful enough to overcome the grip. Well, at 5:32, the longitudinal acceleration (yellow) is 1.11G, slip ratio (red) 7%, and throttle (green) 60%. If we look at 5:35, longitudinal acceleration is 1.22G, slip ratio 5.4%, and throttle 100%.

The slip ratio was taken at the rear wheels since we have a RWD car. Still this is strange. The TC doesn't allow a longitudinal acceleration of more than 1.11G since it closes the throttle, but 1.22G is fine! Now that's some bizarre logic, isn't it? Worse is to come. At 5.37, where we are still accelerating hard, 0.78G, the slip ratio is now -1.9%, which means the rear wheels' longitudinal velocity is slower than the car's velocity, which would implicate that the car has either to slow down, or the front wheels are accelerating the car. Weird. A pattern that seems to appear is that once past 1.15G of longitudinal acceleration and 8% of slip ratio, the throttle is shut, but this is NOT a general rule, and seem to change with gears.

My target is to understand and be able to predict how the TC works in this game. I took a very simple example here: straight line, no bump, and yet it is not straight forward, so what the developers have done to calculate it? I will perform more tests with traction control set at 6 and then 10 just to see.

Give your thoughts!

[Blanchimont]

I remember the day i got GT1, the first thing before i played the game was to read through the complete handbook that was included. I loved the technical descriptions and of course the graphics of the game were windblowing

Thoughts on your TC analysis:

- Please label the axis in your GUI
- Play around with the weight and distribution and see if this changes the longitudinal accelerations as in reality
- Could it be possible that the turbo influences the throttle response and power output in the game? Maybe use a powerful non turbo car in addition
- Use different tyre compounds and aerodynamic settings
- Is tyre temperature modeled and does it influence the game physics?
- Is it possible to read out the tyre normal force?

[mep]

In yellow you see the longitudinal acceleration which should remain constant at a max value corresponding to the grip available, until the engine is not powerful enough to overcome the grip. Well, at 5:32, the longitudinal acceleration (yellow) is 1.11G, slip ratio (red) 7%, and throttle (green) 60%. If we look at 5:35, longitudinal acceleration is 1.22G, slip ratio 5.4%, and throttle 100%.

IMHO, that is a misconception. As long as the engine provides more force than the tires can transmit, acceleration should raise slightly with speed because the downforce increases and so does max transmittable tire force. At 5.35 max engine power is utilised but the aero effect has increased the normal load on the tire.

[MadMatt]

I remember the day i got GT1, the first thing before i played the game was to read through the complete handbook that was included. I loved the technical descriptions and of course the graphics of the game were windblowing

Thoughts on your TC analysis:

- Please label the axis in your GUI
- Play around with the weight and distribution and see if this changes the longitudinal accelerations as in reality
- Could it be possible that the turbo influences the throttle response and power output in the game? Maybe use a powerful non turbo car in addition
- Use different tyre compounds and aerodynamic settings
- Is tyre temperature modeled and does it influence the game physics?
- Is it possible to read out the tyre normal force?

1) I will label everything in the final version, this is a pre-pre-alpha, just to show the general design and the state of the project
2) Gonna add that to the list
3) The turbo pressure value cannot be seen, nor the scale of the gauge, but as it can be predicted, once on boost, the pressure varies with the throttle reduction applied by the TC. I will test NA cars.
4) This is on my list as well
5) What you see in the game is that when you spin a tyre, it color (at the bottom left of the screen) changes towards red when fully spinning. I do not think temperature is modeled in the game, but tyre wear is.
6) No.

IMHO, that is a misconception. As long as the engine provides more force than the tires can transmit, acceleration should raise slightly with speed because the downforce increases and so does max transmittable tire force. At 5.35 max engine power is utilised but the aero effect has increased the normal load on the tire.

You are right here, I might have jumped on conclusions a bit too early, of course the aero load changes the grip available, as well as weight transfer, haven't forgotten that, and since we have a RWD car with big aero loads, it will help. I will take this in consideration in my next tests with other cars.

What would be interesting is to get the instant engine power. There is power graph available with each car, but it is very simplistic (check the bottom left graph):



I will now perform more tests with different cars and see what more we can get about the TC management in the game!

[MadMatt]

I had few more minutes so I gave it a go with a TVR Cerbera Speed 12, a very powerful British car with the following specs:

- Power: 935bhp @ 7600rpm
- Torque: 975Nm @ 5800rpm
- Weight: 943Kg

Fitted in the game with what is called "Sport-Hard" tyres, I performed acceleration tests from 0 to 300km/h with different TC settings, and here are the results:

Code: Select all

TC   Times [s]         TPS average [%]
00 - 16.080 & 15.806 - 98.9 & 98.9
01 - 15.027 & 15.079 - 74.5 & 74.5
03 - 15.043 & 15.048 - 74.4 & 74.4
05 - 15.043 & 15.074 - 74.4 & 74.3
07 - 15.041 & 15.064 - 74.4 & 74.4
10 - 15.038 & 15.043 - 74.3 & 74.3

So what is clear from that small test is that in a straight line, no matter what TC you have, the acceleration is the same!

In fact, the TC does not allow more or less wheel spin to happen. All longitudinal acceleration values were of course stuck at 0.71G when TC was operating, before dropping off as expected. The downforce could be adjusted on the car, and I set it to its minimum value (which was still something) so it wouldn't interfere too much with the results.

Obviously the TC is coupled with lateral acceleration and more probably steering angle as it performs quite well in corners, with the engine being cut more or less depending on your TC setup but it is weird that in a straight line it doesn't respond correctly. Not that surprising when, as I wrote before, you look at the speed of the wheels and that at some point (around 5:36.5 in my first post) the rear wheel speed is lower than the front wheel speed giving incorrect slip ratio values. No wonder they don't work with the slip ratio to define their TC, which is not what we would do in the real world.

I didn't want to criticize Polyphony Digital's work (they might be reading this, who knows), as they are talented and have to make a game that is beautiful, computationally efficient, realistic, so shortcuts have to be made somewhere, but I just wanted to get a bit more to the core of the physic model of the game.

Don't know how much time I want to spend looking at GT6's physics model, but I will continue a bit. See what coast down tests can give me, try to figure out rolling resistance model used in the game, assess grip differences between the tyres, etc.

[Blanchimont]

TPS is throttle position averaged over what time frame?
What was the slip ratio compared to those of the Nissan?
Can you test the same in wet conditions?

[eyalynf1]

I had few more minutes so I gave it a go with a TVR Cerbera Speed 12, a very powerful British car with the following specs:

- Power: 935bhp @ 7600rpm
- Torque: 975Nm @ 5800rpm
- Weight: 943Kg

Fitted in the game with what is called "Sport-Hard" tyres, I performed acceleration tests from 0 to 300km/h with different TC settings, and here are the results:

Code: Select all

TC   Times [s]         TPS average [%]
00 - 16.080 & 15.806 - 98.9 & 98.9
01 - 15.027 & 15.079 - 74.5 & 74.5
03 - 15.043 & 15.048 - 74.4 & 74.4
05 - 15.043 & 15.074 - 74.4 & 74.3
07 - 15.041 & 15.064 - 74.4 & 74.4
10 - 15.038 & 15.043 - 74.3 & 74.3

So what is clear from that small test is that in a straight line, no matter what TC you have, the acceleration is the same!

In fact, the TC does not allow more or less wheel spin to happen. All longitudinal acceleration values were of course stuck at 0.71G when TC was operating, before dropping off as expected. The downforce could be adjusted on the car, and I set it to its minimum value (which was still something) so it wouldn't interfere too much with the results.

Obviously the TC is coupled with lateral acceleration and more probably steering angle as it performs quite well in corners, with the engine being cut more or less depending on your TC setup but it is weird that in a straight line it doesn't respond correctly. Not that surprising when, as I wrote before, you look at the speed of the wheels and that at some point (around 5:36.5 in my first post) the rear wheel speed is lower than the front wheel speed giving incorrect slip ratio values. No wonder they don't work with the slip ratio to define their TC, which is not what we would do in the real world.

I didn't want to criticize Polyphony Digital's work (they might be reading this, who knows), as they are talented and have to make a game that is beautiful, computationally efficient, realistic, so shortcuts have to be made somewhere, but I just wanted to get a bit more to the core of the physic model of the game.

Don't know how much time I want to spend looking at GT6's physics model, but I will continue a bit. See what coast down tests can give me, try to figure out rolling resistance model used in the game, assess grip differences between the tyres, etc.

About 14 years ago, a friend of mine had an E36 M3. It was winter in the midwestern USA with snow and ice on the ground. He called me to come and help him out with his car. It was stuck in its parking space unable to move because of the slippery conditions. With the TC off, the wheels would spin and spin. Wonderful sound from the I6, but no transport. With the TC on, the wheel would spin about 15 degrees, then stop. Still no transport. Just too much torque and not enough grip.

The Speed 12 has about 5 time as much torque as the M3 did, and I've tried driving it in older GT versions. It is nigh undriveable on most tracks. My guess would be that a real physics engine would have it sitting and spinning with no TC, and not moving with TC on, just like the M3 on the ice. That would be very much fun for the game player, so I think they probably set a minimum rolling resistance for the Speed 12 or otherwise overrode the real physics engine just so you could drive the thing!

[MadMatt]

TPS is throttle position averaged over what time frame?
What was the slip ratio compared to those of the Nissan?
Can you test the same in wet conditions?

I am not sure of your reasoning behind these questions.

1) Over the time taken to go from 0 to 300km/h, so pretty much the same time all the time.
2) Slip ratio was much higher with the TVR (from 14% to 18% under TC activation).
3) Not on this track, but I could try on another one, the only issue is that I cannot guarantee a flat track, but I will give it a try.

Your second question made me realize there is a part of my previous message that I haven't pasted here, which was about the slip ratio of the TVR. It gives meaningful results, with positive values wherever you would want them for example, which was not the case with the Nissan, so something is messing up with the dynamics of the Nissan. Being the "racing-soft" tyres, the huge aero load, I don't know, but the values didn't make sense.

[Blanchimont]

I just wanted to understand what the results looked like, no specific thinking behind my questions.

I think in this case, a picture is worth a thousand words.

How does the slip ratio change if you add downforce to the TVR, what is the slip ratio at top speed?
You could find a relation between F at the driven wheels and slip ratio through P=F*v at top speed if you want to go further.

[MadMatt]

I see! Here is a picture of that TVR run (a portion of it):



You clearly see here how the longitudinal acceleration (yellow) is limited to a constant value (0.71G). The slip ratio continues to raise from about 14% at 0:56 to 17.5% at 1:03 when the full throttle is applied. By this time we are in 4th gear at 8200rpm so about time to get back some grip.

It is probably clearer with a NA car but we see that the throttle gradually ramps up as the tractive force at the wheel decreases, which makes sense. Once there is no more wheel spin (in the sense that we know it: tyre smoke), the slip ratio falls down to an acceptable 5%. And when under braking, it goes under 0 so we are all good here. Yes? Well, not totally.

Again, and as I said it, the TC value here made no difference, while it should have. Let me tell you why eyalynf1, even if as you said that car is very powerful. Traction control is by definition a tool to control the traction. Traction is the net force that propel the car forward. But how is that car propelled forward? By its wheels, and we all know that in order to create the tractive force needed, a tyre has to slip (some say skid). This is your slip ratio. So if you control the slip ratio you control the motion of the car, right? Yes, this is traction control.

Take a look at these 3 accelerations from the TVR (remember, all pictures can be opened in a new window to see it better):



They have been made with TC at level 1, 3, and 5. There is NO difference (remember you also have the times posted earlier). I don't know what tyre model they are using in that game, at what slip ratio the peak friction coefficient is, but what is sure is that in GT6 AND IN A STRAIGHT LINE, TC just make sure you don't spin your wheels more than "something that I haven't been able to identify yet", but my feeling is that this is STRONGLY linked with lateral acceleration, which will be my next investigation in order to understand how TC is done in that game.

I will take the short Nürburgring circuit (where it turns 180° right after the stadium), and a Peugeot 905 Evo and play with TC settings. Expect some update later this week!

[silente]

Why not doing that in rFactor?

I can ensure you, physics is much much more realistic and can be really used at least to proper train the drivers.
I proved myself how good for a correlation you could have with real data when you input a well done vehicle model.
The tire model is very elegant and flexible, with many similarities to those use by professional lap simulation software, like ChassisSim, for example.

Moreover, the logging data that rFactor produces include many more channels, including many about Loads, Tires etc.

You can read something about what you can do with it when you have access to good/real data in my Blog:

http://drracing.wordpress.com/

There is already a plugin to export data in Motec and to read them and there are plenty of Motorsports real structures who use rFactor to some extent for their simulations.

I am personally working on a project with an LMP2 team and the results look really interesting. Previously i used it to create a F3 vehicle model for a personal Simulator project to train drivers, with an incredibly good correlation, if you think that the software costs 30 euros or so.

I would be happy to support you in anyway, if you like.

[MadMatt]

Thank you for your interest silente (and nice blog, I am gonna give it a go when I have some time, as I see you like to write as much as me!).

I know that you have more freedom with rFactor or GTR2 for example, but I wanted to do something that hasn't been done before (that new Motec function in GT6 is new and there is no "app" such as this one for GT6). I also wanted to make this project rather simple, so the fact that I don't have a lot of information to work with is not a bad thing. Also I am learning new Matlab features with that project (which was one of my goals), such as playing and working with a video in a GUI, plotting all sorts of graphs LIVE. The code is not very long, around 300 lines at the moment, and it won't exceed 500 I think so small project. I have also worked on improving the computational power my script uses, and I managed to reduce it by 70% which is also something I have never really done in the past.

The good thing with that little project is that I could maybe use some of it for my laptime simulator which I think will interest you a bit more: http://www.f1technical.net/forum/viewto ... 14&t=20639

I have done a bit more progress in understanding GT6's physics model and I will keep working on it before I post something lengthy and annoying here!

[silente]

Matt,

thanks for the link.

I follow your thread already, very interesting! Are you going forward?

I started doing some lap simulation and acceleration simulation myself as well, but i mainly use Excel and VBA for the time being, since my Matlab skillls are very poor.

I am anyway more focused on some Yaw Moment Diagram studies recently (and of course, on driving simulations, as i mention in my Blog). I will write about them soon.

Anyway yes, i like to write very much, although probably what i write could look basic for many of this forum users. But still i think that writing about a topic and sharing is already a very powerful mean to check what you have really understood and to eventually receive some useful feedback.

[MadMatt]

Thank you for the interest! I am having a break from the simulator, as I am stuck and sometimes it is good to take a break, hence this small project! I am not rushed by any team or company to finish my lap time simulator so I take advantage of this hehe.

You should really try to learn Matlab, it is so powerful. There are some good PDFs around to get you started, and good books as well. But once you discover its power, Matlab will eat your soul, like it did to all of us!

Don't worry aboud writing, I think it is a good skill to have. English is not my native language either so it helps. Plus, as they say, sharing is caring!

Regarding this little project, I am working now on improving the GUI, I will post a picture later today, and also improve the track map creation. As we know, drawing a good track map from recorded data is a bit of a black art!