Car Physics for Games (2003) | Hacker News Reader

Car Physics for Games (2003)

asawicki.info

81 pointsibobev13 comments

Timberwolf on Youtube has a series of great videos about car physics in games.

How it developed over the years: https://youtu.be/_IMN9XVYSiY

How a few games worked: https://youtu.be/hcmSBwLKVOY https://youtu.be/p5s-zbXtDoo

Absolutely recommended!

That brings back memories. I remember using that article when creating a simple racing game for a school project in 2005. Using that game I found that the car I was using for a starter series in rallycross at the same time had higher top speed in 4th gear than 5th because of a lack of power

See flight sim book counterpart called “Flights of Fantasy”.

I’m just gonna leave this here …

https://youtu.be/Tn04UwfKk9o?si=8V0kGMSVoBfkZaBT

I’ve lost hours to this.

YMMV

Depending on the velocity settings of your reality

The torque curve per gear is so cool to see... Back in the days I did record my car's engine accelerating on a flat street (and I did it both ways) and then used a fast Fourier transform to find the engine RPM from the recorded sound.

I then plotted that torque curve (it requires knowing how many cylinders, gear ration, wheel size and a few other things and there are a few approximation [like tires deforming with speed, loss, ...] but it's doable) and, sure enough, the resulting curve was very close to the one given in the car's official book.

Yup, back in the eighties car manufacturers (at least some of them) would give you a book with the gear ration and torque curve nicely plotted.

IIRC you only need to record the acceleration over one speed (say the 2nd gear) and then with the gear ratio you can plot all the gears.

It was my first Java app with a GUI! I probably still have the code somewhere on obscure backups.

A lot of the games are currently getting throttle mapping to RPM wrong. In Assetto Corsa if you put 20% into the throttle in neutral the RPMs will climb to red line and the throttle position just determines how quickly that happens and its because they aren't doing this correctly.

Niels Heusinkveld on YouTube has great videos about this!

Just looked this up to see an example of the behavior described. https://www.youtube.com/watch?v=0SlUlENAggE

As a side note, BeamNG.Drive has the most accurate throttle response and audio response of any driving game I've ever tried. You can almost feel the car pull vacuum (or build boost).

You need to simulate the torque curves

You also need to simulate load on the engine even in neutral. With no load, the speed genuinely would rise to red line at a rate controlled by throttle position.

Is this not more so a modern<->old car neutral ethrottle thing? I haven't driven a cable operated throttle car but on bikes 20% will probably hit limiter.

I'm gonna try it after work.

The only precision revving in neutral I've done is for emissions testing a couple times but the throttle tube was barely rotating to keep it at 4k. It was still at that point where the cable just gets tensioned past the freeplay and it was kinda tricky to keep it there since very small movements would move 0,5-1k RPM.

PS: Treating 20% as 20% throttle tube rotation, not necessarily linear to how much chooch you get due to how throttle bodies open.

That's cool. I tried to write a driving game back in the late 80's ... a little like Spy Hunter if anyone remembers that arcade gem. I did lose myself trying to simulate the gearing — got caught in the circular logic problem described in the post.

I wanted to simulate shifting too soon and lugging the engine (vs. shifting at too high an RPM and having thrown away efficiency/performance). But determining when you were lugging the engine depended on the engine RMP which needed to allow for the gear you were in and the current forward speed ... forward speed determined by what gear you were in, what RPM ... so we're back to RPM... Never mind possible slipping of the tire if you broke traction and spinning out or drifting....

Happy to see someone lay it all out in a blog post.

(And happier still that the folks that did the game Carmageddon did a convincing enough job of all the car physics to entertain me for hours back then.)

Im really blown away by the $15 butt dyno phone apps able to pretty accurately measure power/torque of a car based on accelerometer reading and weight/tire size/temp/pressure entered by the user.

Reply on news.ycombinator.com