|
Draco18s Majestic
Registered User
Join date: 19 Sep 2005
Posts: 2,744
|
09-26-2005 16:22
I noticed today that there is a BIG problem with physics engine.
Objects that are not perfect boxes (or spheres?) have improper edge detection, in other words, things don't bounce off of it in the right place, wrong direction, or they bounce too soon (it has been assumed by someone at Linden that the pysics engine calculates the edge as a box surrounding the prim).
I could get a ball to get stuck in midair between two cylinders that were about 3 times the width of the ball apart from each other.
Or floating above a cone.
On further expirimention with my failure at using the physics engine to make a roulete wheel, I discovered that a ball of .02 meters in diameter could get stuck above a box. An ordinary BOX. It sits about .1 meters above the box and fails to fall in any direction (this distance is the same as I found with cylinders). See attached snapshot. The stick between the roulete "wedge" and the sphere was added after the collision, it is exactly .1 meters tall.
|
|
Torley Linden
Enlightenment!
Join date: 15 Sep 2004
Posts: 16,530
|
09-26-2005 16:30
I thinks this is a prob in part because of Havok 1 limitations, bounding boxes and imprecisions in the physics engine. I know there's someone here who can give a more technical answer.  |
|
Ardith Mifflin
Mecha Fiend
Join date: 5 Jun 2004
Posts: 1,416
|
09-26-2005 16:31
From: Draco18s Majestic I noticed today that there is a BIG problem with physics engine.
Objects that are not perfect boxes (or spheres?) have improper edge detection, in other words, things don't bounce off of it in the right place, wrong direction, or they bounce too soon (it has been assumed by someone at Linden that the pysics engine calculates the edge as a box surrounding the prim).
I could get a ball to get stuck in midair between two cylinders that were about 3 times the width of the ball apart from each other.
Or floating above a cone.
On further expirimention with my failure at using the physics engine to make a roulete wheel, I discovered that a ball of .02 meters in diameter could get stuck above a box. An ordinary BOX. It sits about .1 meters above the box and fails to fall in any direction (this distance is the same as I found with cylinders). See attached snapshot. The stick between the roulete "wedge" and the sphere was added after the collision, it is exactly .1 meters tall. Welcome to the world of shitty physics. Eventually, they'll be upgrading to Havok 2 which promises to improve things considerably. However, that upgrade has been "just around the corner" since I joined last year. I think the next estimate is sometime early next year or three days before Ragnarok. My money's on Ragnarok. |
|
Draco18s Majestic
Registered User
Join date: 19 Sep 2005
Posts: 2,744
|
09-26-2005 16:33
Kinda what I figured. But with a box? Man...shity physics indeed...
|
|
Lee Linden
llBuildMonkey();
Join date: 31 Dec 1969
Posts: 743
|
09-29-2005 10:27
To my knowledge, Havok 1 creates a bounding box 0.10 m around the physical object, which is then used for determing physics. It does match the shape of the object (i.e. your theory that the sphere is colliding as a cube is incorrect), but it means you can't effectively perform high-collision physics with minute objects under the current release.
|
|
Draco18s Majestic
Registered User
Join date: 19 Sep 2005
Posts: 2,744
|
09-29-2005 10:29
I'm not sure that it is the shape of the object. Otherwise I could get a ball to roll down the side of a cone, when clearly, I can not.
Ah, well, I pray that 2nd Life get upgraded to Havok 3 soon.
|