Real-Time Physics Simulation Forum

i detected that raycasts and contact tests in bullet are not working well when rigid body is moving relativly not too fast. And this is not problem of rendering/interpolating. You can observe that in bullet demo box2d by: 1. picking body 2. throwing it to set it some velocity 3. pause simulation (ke...

Ok, I have done that by way around, copying contacts every physics tick to vector in my wraper on btRigidBody after that i can look only on contacts that i am interested in (for specyfied body). This works lot faster than doing contact tests. for (int i=0;i<PhysicsObjects.size();i++) PhysicsObjects....

Yes, i now that raycast is better for bullets. I called it wrong. I meant slow moving rigidbody which must now about itself collision. For example,granade or spaceship which fly(via applying forces) and if collide with player kills him, raycast can't be solution for that

i was thinking about creating new colision object which would behave like simple rigid body but in every step simulation will store contact points which concern this body. It would be like hybryd btRigidBody and btGhostObject. Why this may be needed to? Example: bullet(small rigidbody) from weapon h...

i think the best way to achieve this is to iterate over all dynamic bodies and apply central inpulse/force in the direction that you want, on every physics tick, like this: int i; for (i=dynamicsWorld->getNumCollisionObjects()-1; i>=0 ;i--) { btCollisionObject* obj = dynamicsWorld->getCollisionObjec...

It is better to register contact callback, for body or call contact test every physics tick to improve performance.
How much worse is calling contact test every tick, or better, or performance is the same both methods?

is this possible to create in bullet engine, concave 2d static collision object? I have tried with btBvhTriangleMeshShape with z-coordinate = 0, but box2dshape don't collide with it. btConvex2dShape works fine, but it would be great to create concave. Any ideas?