The CollisionGeometry::computeMOmentofInertiaRelatedToCOM is naively implemented. It explicitly calls the virtual methods to compute moment of inertia, center of mass, and volume. For simple primitive shapes where these quantities are O(1) operations that's not bad. But for meshes this is horrible. These quantities are so tightly coupled that it makes sense to have the option to compute them simultaneously (to reduce the cost of the integrals).
I propose that the CollisionGeometry class be given a new virtual method:
/// Compute the volume of the geometry while simultaneously reporting the center of mass and the inertial tensor in the world frame (returned in the given out parameters).
virtual S computeMassProperties(Matrix3<S>* inertia_tensor, Vector3<S>* com)
It's default implementation can be exactly those three calls currently being made, so most geometry sub-classes can ignore it. But more complex geometries can override it and solve for all three quantities simultaneously.
sherm1
commented
6 years ago
The CollisionGeometry::computeMOmentofInertiaRelatedToCOM is naively implemented. It explicitly calls the virtual methods to compute moment of inertia, center of mass, and volume. For simple primitive shapes where these quantities are O(1) operations that's not bad. But for meshes this is horrible. These quantities are so tightly coupled that it makes sense to have the option to compute them simultaneously (to reduce the cost of the integrals).
I propose that the
CollisionGeometryclass be given a newvirtualmethod:It's default implementation can be exactly those three calls currently being made, so most geometry sub-classes can ignore it. But more complex geometries can override it and solve for all three quantities simultaneously.