Real-Time Physics Simulation Forum

If you're doing Deep Robotic Learning (DRI) and Human Robot Interactions (HRI) then you probably want to be using pybullet and asking your questions in the pybullet forum channel.

The statement: "when I add a polygon to a btTriangleShape" does not quite make sense. A btTriangleShape is just a triangle with three vertices and doesn't have polygons. Perhaps you meant a btTriangleMeshShape? btTriangleMeshShape is a pure virtual base class. The first non-virtual derived class is ...

It has been reported in previous forum posts (which I wasn't able to find with a quick search) that an object will sometimes NOT bounce on collision when the substep is too small. Since I couldn't find those threads I have to rely on my memory which tells me the problem was seen at substeps of 1/100...

This might affect things, you'll have to test: I don't think your world->stepSimulation() is doing what you think it is. You're specifying the "step" but not the "substep" duration and hence it uses the default substep of 1/60 sec. So your world "steps" forward in 1/600 sec increments and does nothi...

I dunno the answer but I have some questions about your simulation: (1) What was your simulation stepping strategy? Fixed substep? If so, what was its duration? How man max substeps? (2) Are you using single or double precision? (3) Your ball was an implicit btSphereShape , I assume, but what shape ...

I started looking at the btSoftBody code to see if I could provide some insight. I don't fully understand it yet but it looks like btSoftBody::m_faces is a btAllignedObjectArray<btSoftBody::Face> and a Face is a "triangle" that has pointers to three Nodes . struct Face : Feature { Node* m_n[3]; // N...

The linker can't find the file: /usr/lib/gcc/x86_64-linux-gnu/libdl.so Look for it to see if you have it. The file path relies on you having a directory called: / usr/lib/gcc/x86_64-linux-gnu/5/ Do you have that directory? That appears to be an old version of gcc . On my linux machine I see 6 and 7 ...

After a bit of cleanup for readability your old code looks like this: btVector3 pivotAnchor(0,0,200); btVector3 pivotSpinner(0,0,200); btVector3 axisAnchor(0,1,0); btVector3 axisSpinner(0,1,0); btHingeConstraint joint = new btHingeConstraint(anchor, spinner pivotAnchor, pivotSpinner, axisAnchor, axi...

It would be helpful if you were to supply source code. It is hard to make sense of what you are doing wrong from your description, however I will hazard some advice: From your description it sounds like you are using world-frame coordinates. However the transforms in the btHingeConstriant constructo...

The btCapsuleShape is inside the BulletCollision library. We must conclude the linker is unable to find the library itself.

That old thread is still correct except for one thing: Using btCompoundShape is not the only way to go. The OP might have been able to use Bullet's GImpactShape stuff which allows for dynamic convex shapes. I would sum it up like so: (1) btTriangleMeshShape can NOT be used for dynamic objects. The r...

The little boxes are using btTriangleMeshShape? Or just the big cyan container? Bodies that use btTriangleMeshShape cannot be dynamic else Bad Things happen.

Now that I look back at this issue I notice: it doesn't look like the OP was creating the motionState and then not using it. The motionState is a "bridge" object between the btRigidBody and the GameObject (rendered thing). One should at least supply the motionState to the btRigidBody constructor, an...

I don't know how to do it exactly but merely mention it as a possibility. You'd have to identify the GJK parts, study its API, and figure out how to make it do what you want. Dunno how or if Bullet's GJK algorithm can be fine-tuned.

How do you extract the worldTransform of the RigidBody ? Are you using something like DefaultMotionState::getWorldTransform() ? If so then the following issue may apply: Bullet's MotionState API will supply an extrapolated transform to the MotionState rather than the RigidBody 's true transform at t...