Newton box falls straight through the mesh collision

[Neo]

I use plane to build as a tree collison, but when I create a Nowton box into the world, it falls through the plane,
as if the plane is not exist...
Then I tried hard to find solution, it seems that mesh tree collison only works when the mesh is a convex hull,am i right?

[Julio Jerez]

not collision tree, works with any kind of polygons that do no self intesect: convex or concave.

[PJani]

Are you setting inertia?

[Neo]

not collision tree, works with any kind of polygons that do no self intesect: convex or concave.

still cannot work..then i change the terran model into a newton box...this time collison happens ,but sth wierld took place:the box i dropped from a very low place bounce up very high,and had fly out of the newton world directly(normally the box fall down and disapear when hit the floor, then i set the gravity and mass to a very small value, such as mass=0.1 and gravity=0.01. then i found that it's not disapear but bounce very high, even higher than the original drop hight).....

ps:
i use lrrlicht 1.7.1 for graphic perpose and
my newton version is 2.35, download from googlecode.

[Neo]

Are you setting inertia?

of course, use the NewtonConvexCollisionCalculateInertialMatrix() and NewtonBodySetMassMatrix()

[Julio Jerez]

get version 3.00 for svn

[Neo]

get version 3.00 for svn

NewtonSetWorldSize and NewtonReleaseCollision have disapeared in the v3.00...
Is it a change to old version?

[Neo]

get version 3.00 for svn

same result...

[Neo]

I let the console print out veloclity every time when game update, and then i found out that forthe first few seconds(my gravity is -0.01,and mess is 1 , so i can observe it clearly), the box accelerate like a normal stuff falling, but then the velocity start reduce. After hit the floor(a large newton box with 0 mass), it bounce at amost 10 times as the falling speed and fly to the sky....then it velocity began to reduce to minus, box start to fall. However, the falling speed still repeat the accelerate-for-shot-time-then-reduce's process...WTF....

[Neo]

well, i've wrote this program by modifing this examlehttp://newtondynamics.com/wiki/index.php5?title=Tutorial_-_Irrlicht_and_Newton_start, later i found that collision works well when i change the getRelativeTransformation to getAbusoluteTransformation, though the poision isn't right(the model came back to thr origin)...but i believe i can handle it , just adding some code to modify the posion.
but the point is, now the floor(the huge newton box), when i tried to load it as tree collision, the box i dropped still fall through it:(
I want load a tree collision as terrain and use bouding box of model to deel with collison, briefly speaking, a tree collison with multiple primitive collison. How can i get it work?
BTW, does tree collison and convex hull collides each other? thx

[Julio Jerez]

yes everything collide with everything else, with execption of static mesh versus static meshes (collision tress vs collision tress or hieghtfield)

[Neo]

yesy every thing collide with every thing else, wit th exemtion of static mesh versus static meshes

thx, the convex hull works, but collison happens before the scene node actually hit (they are far from each other)
it seems that the newton body's location is differ to the irrlicht scene,my function code to create the convex hull is quite familiar with ithttp://newtondynamics.com/forum/viewtopic.php?f=9&t=7230&start=30, what did i do addtionally is to create a newton body with that collison object.
and transform it

Code: Select all

NewtonCollision* collision_obj=NewtonCreateConvexHull(nWorld, nVertices, &vertices[0], sizeof(f32)*3,0.001f,10,NULL);
 
                matrix4 matrix=irr_meshS->getAbsoluteTransformation();
                vector3df origin =matrix.getTranslation()*IrrToNewton;//irrToNewton is set to (1/32)
                matrix.setTranslation(origin);
                NewtonBody* cb=NewtonCreateBody(nWorld,collision_obj,&matrix.pointer()[0]);
 

is it the transform cause the problem? i know nothing about transform matrix

[Julio Jerez]

if you are using Netwon 300, you cna us eteh serialization and test teh scene in teh sandbox demos

all you nee to do is this.
after you create the workd or at any time you can call this funtion

void NewtonSerializeToFile (const NewtonWorld* const newtonWorld, const char* const filename);

then you can load the file in the sandbox and see if they are differences between what you see and what you the world is doing.

[Neo]

if you are using Netwon 300, you cna us eteh serialization and test teh scene in teh sandbox demos

all you nee to do is this.
after you create the workd or at any time you can call this funtion

void NewtonSerializeToFile (const NewtonWorld* const newtonWorld, const char* const filename);

then you can load the file in the sandbox and see if they are differences between what you see and what you the world is doing.

thx a lot Julio, it do help a lot.
what i'm seeing is, my tank box apear in the right place with right scale(because it's a primitive collison, not user mesh), bu the scene(this time only a larger box, i use it as floor, but load it as user mesh collison)become very large. to the degree that when i start the scene, i am in the floor's collison box! It's a very huge box, ten times larger than the actual size of the scene node. yeah, the it's the problem of scaling, maybe i should scale the point when i add faces to the tree collision or transform vertex when i load them to the convex hull?

PS:Did the NewtonReleaseCollison have removed from Newton 300? The compiler said there's no function called NewtonReleaseCollison..

[Julio Jerez]

NewtonReleaseCollsion is deprecated.

you can try using the scale feature of the collsion to match the size of you visual mesh