My wish for Channel 9 to augment the trademark videos with downloadable audio files has been granted. As a result, I was able to listen to Charles Torre’s interview with Brian Beckman during my exercise hour yesterday. Even if you’re not a gamer (I’m not), there’s a good chance you’ll enjoy Brian’s lively explanation of the physics that govern the simulation of planes and cars. I guess I’d been vaguely aware that PC flight simulators predated PC racecar simulators by many years, but I’d never thought about why.
To do a credible simulation of a plane, Brian says, you only need to account for two coordinate systems (the earth’s and the plane’s) and four forces (lift, drag, thrust, gravity). Because you can do table look-up for the values of variables such as lift and drag, and because those tables are small, you don’t need a “fast and fat” computer (fast CPU, big memory) to do the job. That’s why Flight Simulator was possible on the earliest PCs.
Since only some of us fly, but most of us drive, you might have expected automotive simulators to have arrived sooner than the mid-1990s. And they would have, were it not the case that car physics is, counter-intuitively, way more complicated than airplane physics:
If you analogize a tire to a wing — that is, the thing that generates the force — you’ve got four of them. And they’re connected to the car by these complicated linkages. Maybe they’re McPherson struts, maybe they’re independent suspensions, maybe they’re leaf springs. All of these things act differently, so each different car is going to have not just different shades of physics, but completely different kinds of physics, completely different equations.
It goes on from there: more coordinate frames to account for, complex dynamics of steering, acceleration, and braking. It’s fascinating to learn why accurate simulation of cars only recently became possible on PCs. Along the way, you’ll be reminded about the properties of the safety envelope that ordinary drivers seldom push but that race drivers always do.
The interview concludes with another counter-intuitive observation. Although the ability to crunch complex nonlinear equations at 30 or 60 frames per second is what makes these simulations possible today, Brian thinks this radically different approach is the way of the future. The Rigs and Rods system shown in that video, created by Pierre-Michel Ricordel, uses a technique that Brian calls “sticks and stones”:
Stones, which have mass, and sticks between them which are little flexible things we can model with very simple physics: harmonic oscillators. That’s all you need, just one physics model, a damped harmonic oscillator connecting pairs of stones, and, by gum, you can simulate really amazing things.
The computer is now fast enough to be able to simulate hundreds or thousands of these independent systems simultaneously, every step. We can now dispense with the really hard mathematics. If you’d asked me to bet money that this were even possible, I’d have said no. But that’s because I had this long experience where you just didn’t think about doing thousands of particles iterated in a system like this. Here’s a guy who took a fresh approach, he said let me see what I can do, and sure enough, this is a magnificent system.
We’ve heard it before, we’ll hear it again: a network of many simple parts trumps one big complex monolith. It’s a story that keeps on surprising us, but probably shouldn’t.
So, you might wonder how a video with as much visual content as this one — equation-filled slides, Mathematica screens, YouTube videos — fares in audio format. In this case, surprisingly well. It’s amazing how much information voice alone can carry. Later I did review parts of the video, and having listened I knew just which parts I wanted to see. But this exercise confirmed my belief that downshifting from video to audio is a really useful way to give people access to long-form material they might not have time to watch.