Grey

If you're not using any soft bodies, then in theory your first bullet point should be correct. But we've seen cases where things don't work out that way because of...

That's a very interesting outcome. I vaguely remember an issue with FCL where we originally used a map or set of pointers to store the collision objects that would get...

I don't think this specific header was ever meant to be part of the public API. DART has just historically not made a hard distinction between public and internal API...

To test a hypothetical `0.4.2` release of rendy, you can use the following patches to the current master branch of amethyst: ``` diff --git a/Cargo.toml b/Cargo.toml index a3cdf092..179e0994 100644 ---...

I think having an optional constraint solver that's higher performance but maybe less feature-rich is a great idea. The biggest catch is that it might be difficult to communicate the...

> Does this imply that moving the joint values to the skeleton? I was mostly thinking of all the cached `Eigen::VectorXd` data structures. It will be significantly more difficult to...

I think rather than trying to "centralize" things into `Skeleton`, we'd probably be better off devising a custom allocator that can keep the memory together.

Well not so much "devising" a custom allocator, but enabling the dartsim API to accept a custom allocator.

If you mean closed kinematic chains, there is some limited support for them. Kinematic chains can be closed by applying a joint-like constraint between two bodies. Currently there are two...

Correct, not currently. However, it may be possible to do a workaround where you have a dummy body attached by a normal hinge (revolute) joint and then close the loop...