Wednesday, 24 April 2013

rFactor 2 and the new tire model - First impressions

It's generally accepted that one of the most important elements (if not, the most important element) in any racing sim is the tire simulation. People often say the ultimate performance of a car comes down to what's going on with the contact patches in each corner, which is what makes simulating them well so important.

I can't deny that in terms of technical comprehensiveness, the tire model in rFactor 2 is much more comprehensive than it is in rFactor 1. And I have learned a lot from it, it's been fascinating to work with. But it seems inappropriate for modders.

One element of building an rF2 tire is the design and construction of the tire. Different ply layers, different materials on the different layers, different tread depths... And the overall shape, as it is when the pressure inside is the same as on the outside. After that, you can see how the tire deforms under different conditions. How the middle bulges out with pressure differences and speed, how it compresses under loads, stuff like that.

A lookup table is built from this construction, and that's then combined with some more parameters in real time in engine. Here's where you control things like how grip builds up until the peak slip angle, how it drops off beyond there, that sort of thing. This mix of offline simulation and realtime simulation makes for a very comprehensive result, more so than a 100% realtime model.

Problems arise quite quickly with this solution, though. Building that lookup table in a pretty low resolution (for prototyping) takes about 4 hours per tire on my PC. And that needs to be repeated for the rear tires too. Ideally, it is repeated for every compound aswell (because softer compounds are literally softer). So, prototyping these tires in engine is a long process.

It's also difficult to analyse what you've got. There's a lot of mystery and confusion around the parameters we can change, and it's also not 100% clear where the lookup table ends and the realtime model begins. The tire testing tool provides an environment where you can subject the tire to any combination of forces you like, but it won't plot the outputs onto a graph. Instead, all you get is numerical outputs which fluctuate. So if you want to try and plot out slip curves, load sensitivity, etc. You're doing it by hand for the glimpses you get as it fluctuates. This is -not- an efficient way to confirm your work is correct.

However, what I think is the biggest issue with this tire model is the fact that we need to include the construction of the tire on most (or all, depending on the developer's choice) layers. This isn't information that's readily available. On top of that, each material's elasticity, thermal conductivity, etc. must be described too, as well as how these things change with temperature. This too is information that is not easily available. As far as I can tell, most people releasing rF2 cars have used the ISI tires as reference, or even as a base.

rFactor 2 certainly hasn't seen the adoption we were all hoping for. The cars available for it are pretty limited, and from what I hear, the tire model is the most common complaint from modders. It's certainly putting me off, and I certainly won't be taking on any more rF2 projects until I at least have a faster way of designing these tires.

No comments: