Real-Time Physics Simulation Forum

Ja, this game looks amazing...

...and even has an editor mode where you can build levels and share them (first video in the following link)

http://www.gametrailers.com/gamepage.php?id=4491

The physics look great, but the gameplay and visual design make it just perfect.

There also seems to be a lot of work in context aware character animation, (characters hanging on gear's, running on balls).

To be honest, It's kind of "the game I've always wanted to make", so I'm late (again)

Oscar

Oscar Civit Flores wrote:Ja, this game looks amazing...

...and even has an editor mode where you can build levels and share them (first video in the following link)

http://www.gametrailers.com/gamepage.php?id=4491

The physics look great, but the gameplay and visual design make it just perfect.

There also seems to be a lot of work in context aware character animation, (characters hanging on gear's, running on balls).

To be honest, It's kind of "the game I've always wanted to make", so I'm late (again)

Oscar

I've to agree, totally...

Carlo Lanzotti

This game is the proof of the idea of physics based game design. Alas, now I might have to buy a PS3.

In this interview about LBP:

http://www.gamasutra.com/php-bin/news_i ... tory=13051

I found the following observation quite interesting:

Concepting was also a large part in resolving what the company had termed their "screwdriver fights," settling, for instance, an early battle about how a 2D physics engine could function in a 3D world.

I've read something similar in an interview with the developers of "Steam Brigade" (which used Newton, I think)

I have plans for a personal project in the same lines but with different gameplay, so I wonder... which is the best way to cast a 3D physics engine to work in 2D?
I've come up with some "rules":

• - Add a planar constraint to all objects' CoM.
  - Add a "Hinge" constraint to all rigid objects.
  - Filter collision normals, user-forces/impulses, and, in general, any vector magnitude that may not lie in the YX plane.

Would you add/remove/modify anything?

How about building an specialized 2D API on top of the actual 3D engine that performs all tests, projections and (implicit) constraint setup? Is there anything similar "out there"?

Regards,

Oscar

PD: I'm going to buy a PS3 too, but living in the "Old Continent" I'll have to wait... anyhow, LBP is planned for 2008 and there's no other game I'd do it for, so there's no hurry

Oscar Civit Flores wrote:

• - Add a planar constraint to all objects' CoM.
  - Add a "Hinge" constraint to all rigid objects.
  - Filter collision normals, user-forces/impulses, and, in general, any vector magnitude that may not lie in the YX plane.

The planar and hinge constraints are interesting ideas, but the cost would probably add up quickly. Giving the rigid body's intertia tensor infinite values in the XY plane (or more likely, zero values for the inverse inertia tensor) should ensure that it only rotates along its Z axis, although I'd apply some sort of "correction" torque too, just in case round-off error accumulates enough to be a problem.

-Chris

Hi,
I considered the Inv inertia tensor (and inverse mass matrix) trick, but I don't think it's a general and "safe" solution... don't you end up with a singular matrix?

And in any case you can't use the direct inertia tensor, because some elements would be infinity, so if the engine uses it for something else than solving constraints (like computing the angular momentum from a user-given angular velocity, for example), you may get into trouble.

I't true that the planar/hinge constraints would add an overhead, but I don't think it'd be too noticeable, given that they're objecto-to-environment constraints which could be solved locally for free objects.

Regards,

Oscar

Oscar Civit Flores wrote:Hi,
I considered the Inv inertia tensor (and inverse mass matrix) trick, but I don't think it's a general and "safe" solution... don't you end up with a singular matrix?

And in any case you can't use the direct inertia tensor, because some elements would be infinity, so if the engine uses it for something else than solving constraints (like computing the angular momentum from a user-given angular velocity, for example), you may get into trouble.

I guess it depends on your engine. I'm almost positive it will work in Havok (which is the only one that I'm familiar with), but you're right - it will make the matrix singular, but that should only be a problem if you ever need to invert the matrix.

Perhaps they're just using a 2D physics simulation and faking the z component? I'd be interested in finding out how well that works.. it seems like it might be confusing unless players had an explicit "move in/out" control.