Wolfire Games Forums

I am current reading a book called "Game physics" by David H eberly, and the book is quite heavy on math and physics.
Things like inertia tensor all sort of integrals and algebra.
Since david is still a school student, I wonder how did he create the physics engine? did he learn all the undergraduate math and physics? or did he use some cookbooks or "hacks"?

Just wondering.

Good question. I'd like to hear his (or Jeff's) response

The physics engine David is currently working on cannot be found in books. It is based on original and ongoing research by Alec Rivers at Cornell University. How is David capable of understanding it? He is pretty smart.

FastLSM isn't exactly rocket science. Besides, wasn't David involved in research?

The physics in Lugaru are based on simple particle-stick constraints, which are honestly much easier to use (and more robust) than object-level dynamics based on inertia tensors. It is very similar to the physics used in the original hitman games (which they wrote a paper about. It is in a sense the same underlying idea behind Alec's physics engine... break the whole model down into small parts, and then simulate their (relatively) simple constraints. This results in a convincing and stable simulation for the whole complex object.

rudel_ic wrote:FastLSM isn't exactly rocket science. Besides, wasn't David involved in research?

What?

David wrote:The physics in Lugaru are based on simple particle-stick constraints, which are honestly much easier to use (and more robust) than object-level dynamics based on inertia tensors. It is very similar to the physics used in the original hitman games (which they wrote a paper about. It is in a sense the same underlying idea behind Alec's physics engine... break the whole model down into small parts, and then simulate their (relatively) simple constraints. This results in a convincing and stable simulation for the whole complex object.

rudel_ic wrote:FastLSM isn't exactly rocket science. Besides, wasn't David involved in research?

What?

Wow, this paper is a lot shorter then the 700 pages book I am reading.
But of course it is no subtitude for a good theory base.
Thanks, I think I will start implement a physics engine using this article and read the book as well so I will also be able to improove the engine based on theory.

David wrote:

rudel_ic wrote:FastLSM isn't exactly rocket science. Besides, wasn't David involved in research?

What?

Well, he thanks you in his paper here. I thought he did that not just for your pizza delivery

How suitable. Everyone loves penguins.

I helped with some of the graphics in his demo, but mostly somewhat unrelated stuff like rendering images over each other and making the shadows look nice. I am still working on getting the fractured objects to look better by generating 'cap' faces to cover up gaps created in the surface.

Those physics look so nice, and when he showed that baby, I thought they were going to cut him.


The shadows are great, usually designers have to cheat and put in fake shadows that look like the models, but these are perfect. Even when the models switch from mesh to cubic, the shadows follow them perfectly and when the models glitch out with because of linear and quadratic physics, the shadows do the same thing, it's spot on.


This technology is just so awesome, it's fast, you can break it, and it looks amazing. How can so few people, design something so great?

David wrote:I am still working on getting the fractured objects to look better by generating 'cap' faces to cover up gaps created in the surface.

Clone an adjacent face that is preferably orthogonal to the shearing direction and stretch it over the gap, I say.

Edit: May only work for small gaps, though. Split the gap-face and shrink the halves when you really want to rip them apart..?

Whoops, you're probably only using triangles, eh? Hmm.

Fastlsm is different from realmatter's fractal physics? Is it more optimized to run better?


Or is FastLSM the same as fractal physics, and it's already more optimized than regular crate and ragdoll physics?

In either case, sickness.

How would we know? We can read the algorithms for FastLSM, but not for this realmatter thing.

But as I see it, FastLSM is at least more generic than realmatter.

Also, ask David He seems to be pretty involved in this.

In other news, I thought about the cap faces problem, and as it turns out, the problem can be reduced to 4 cases with triangles. Unfortunately, you need an even number of triangles at the gap edges for my approach.

Fastlsm is different from realmatter's fractal physics? Is it more optimized to run better?

Here is Alec Rivers' site with his paper describing his fast lattice shape matching algorithm. FastLSM is the fundamental algorithm behind his physics engine, and Realmatter is his implementation of it. I don't think there are any fractals involved... are you thinking of fracture? The LSM method is unusual in that it still works if you split cells, so at high enough resolution it can support any kind of cut or stress fracture.

BTW, the physics for lugaru1/blackshades/glfighters is a built in ragdoll simulator (and all the problems thatcome with it—jittering, anyone?). Anyway, L2 already looks to pe promising as a phyics engine-check out the teapot!