Posted by: drracing | February 3, 2013

Driving simulation

Here again.

Once more, too much time since the last time i wrote here.

During the last months I have spent much of my spare time working on driving simulation, following what I have already described a bit in my last post. It is really a very interesting world, both from an entertainment perspective (to drive a cool race car that YOU modeled, above all if you modeled it properly, it´s really exciting, although it is only in a virtual world!) and, more important here, from an engineering perspective. To be able to model each component the right way in a driving simulator, first of all you need to know your component very well and then you need to know the better you can how to build it into the simulator, so, actually, what to write into the text files that represents each car area behavior so that the car behaves in at least a similar way to the real counterpart.

Using my Formula 3 model as a “school bench” during the last months I had the possibility to go deeper and deeper into the modeling tricks required to properly “design” each area of the car into the sim software (rF in this case). And this is really like climbing a mountain, a very high one! Sometimes, during this process, you can also see how some of the ways chosen to model some areas of the car are probably not the best the developer could choose, but you have anyway to live with it.

To learn the language you have to speak to model your car is a long but necessary/useful process. And it also forces you to better know the physics behind each component behavior, both to be able to reproduce it and to understand the limitations of the approach you must use. Everything here is represented by one or more equations where you can choose some parameters values to draw the right “curve shape”. This brings you sometimes to face the limitations of the code, that maybe doesn´t allow you to depict some features at all (or, at least, not the way you would need to do to be accurate).  Or sometimes you simply have to admit that you don´t know enough about something and you need to do your homework to be able to model it properly.

I already spoke about tires in my last post and, as I said, I had the possibility to see some of the good and the “less good” things about the way you have to use to model each of their features (slip angle curves, load sensitivity, temperature sensitivity, pressure sensitivity, etc). Sometimes, anyway (but not only here, as I will explain) the problem is also that you don´t really know enough into details how a real tire behaves regardingcertain parameters variations. See temperature, for example, or inflating pressure: which is the optimum grip inflating pressure for a certain tire (maybe easy), how this pressure value change with load and how much? Which is the best performance temperature for a certain tyre, how much grip you loose for, say, 10° more than the optimal temperature or 10° less? How your grip loss evolves with temperature variation? These are just examples of things that, probably, even a tire manufacturer would not be able to model accurately (not all of them at least), but that you need to insert values for into the sim. So, sooner or later, you are forced to make assumptions based on your experience (or on your common sense) and to verify them driving the car, which can anyway be funny! The problem is, in some cases you will never know how much this parameters are far from the real values and probably many people like you will never know. So, if you just care of the simulation from a vehicle dynamics perspective, sometimes you could even decide to kill these parameters influence (to ignore it, making the model in an insensitive way to these parameters). This is of course not nice if you want to develop a racing simulation for the public or for somebody who is interested to see, for example, how tire pressure evolution during a race could influence grip or tire wear.

Anyway, to sum up, now I have a nice set of tools (mainly excel sheets) to extract the numbers I need to put into the rF text files to properly model a car. Of course, it would have been nicer to make something able to also create (or fill) the text file automatically, but my VBA knowledge is practically equal to nothing, so I will accept to copy and paste the numbers manually.
Unfortunately, I didn´t have the time yet to have now some fun testing things on the model: actually I mainly just validated it, comparing with real world results (and getting really good match, this confirming that with real data in, the model perform very similarly to the real one). Now, to be sure that some parameters are really validated, I would need a good driver (or at least somebody better than me) to really push the car to the limit and see if something must be somehow corrected (see mainly tires grip level, which can actually change a lot, also in real racetracks, depending on the conditions). Anyway, a nice and funny consequence of the absence of a good test driver is that I was force to be the driver as well and this has been not only enjable, but also very useful to improve the understanding of the problems connected to the system “car + driver” and to how the driver could influence its behavior/performance. The possibility to check, after each test, what you have done on the track with Data Logging is invaluable on that side, also to improve your skills (exactly the same way you help real drivers in real racetracks when you show them data logged).

It would now be nice to take some time to really and properly test reactions to modifications to the setup/tires/aerodynamics/engine etc. to see both how the models respond and also how the change could affect the performance. One thing I actually learnt during the process is that, in some conditions, a driving simulations could tell you (if you are confident enough to trust it) more about a set up change than what a “performance – lap time simulation” could sometimes do. Ok, we are always talking about videogames, but the presence of a “driver in the loop” actually increase very much the sensitivity to certain changes, mainly because also the driver´s performance will change, as well as his style/way of driving/lines etc. Have you ever tried to change from 0 to, say, 400 N/mm the antirollbar stiffness into a lap time simulation (see Bosch Lapsim for example, as it was free) and see how much lap time change? If you have never done it, you could be surprised to see how small the corners speed and the overall lap time difference is; but if you think carefully to the problem, it should be clear why it is not so strange: this software explores the car limits, or at least the simulated car limits. No fear, no headache, no objects on the track, no bad mood in the morning…But does a real driver explore the real car limit as well? And how a change could be amplified by the fact that after it he likes the car more? Lap time simulations turns to be very useful to see big trends and change to main parameters like weight, downforce, tires grip, engine power, etc., but will often not tell you much about the fine setup change. As I already said, I would not even trust blindly a driving simulator about that (there are so many limitations to be considered, not only in the model or in the code but also, for example, in the way the driver drives the simulated car: is this the same he drives the real one? I don´t think so), probably I would never use exactly the same setup you use in the simulator on the real car. But probably, talking about relatives changes, you could have a better understanding of how changing some parameters could influence the performance. Of course, it is always a rough simulation, not the real world. But you can always learn something out of a simulation, no matter how much you simplify the story. There must be a reason if all top end motorsport teams (in F1, Le Mans, etc) are now relying more and more on the work at the simulator. Of course, we could probably never compare the complexity and the fidelity of their hardware/software to an “house-made” rF model. But the principles are very similar.

Moreover, as I already said, after a session in the simulator, you have the possibility to check with your data acquisition parameters that most part of the race engineers in the world will never be able to acquire, see for example tires forces. Again, it is also in this phase that you can really see the limitations of the code in some aspects…but hey! It is always a video game, right?

Canali motec F3_1 Canali motec F3_2
To close this post, next topic of my new hobby will be LMP cars. I am already at a good point modeling an AUDI R18 in 2011 spec. Of course, i don’t have any personal reference on these cars, i have never worked on them. The process is/will be a kind of reverse engineering one, mainly based on material you can find on the internet, pictures, assumptions and sometimes tyre data from tires similar to the one these cars are using.

By the way, one thing already clear from what i have done till now modeling AUDI R18 is these cars are really amazing!

Audi R18 Silverstone


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s


%d bloggers like this: