Jason K. Moore

@Davide-sd thanks for reporting these. There are several prior issues where we worked out the sign convention ideas for the code. There has been general confusion on how to approach...

If you want the equations to be differentiable, one trick is to replace a Heaviside with a "steep" the logistic function: https://en.wikipedia.org/wiki/Logistic_function I don't think lambdify will work with a...

Otherwise you have to make two models that are each a function of the two heavside cases and then stop the simulation when the heavside transitions across the nonlinearity and...

One more trick. If you use SymPy Abs() to construct a function that is similar to Heaveside `(x + Abs(x))/2/x` as you set assumptions on x being real, then I...

Yes, you may have to use a number very close to zero for the initial condition.

I recommend that you stop integration on impact, calculate pre and post momentum given some coefficient of elasiticty, then start the susequent simulation. For a problem like this, that is...

> Should we support quaternion and dcm orientations? Ideally yes. With only the body/space fixed DCMs gimbal lock is possible. But this is often fine for many practical use cases....

> Could you give an example of what you mean with this? I guess I'm imagining something like: ```python sjoint = SphericalJoint(..., dependent_speeds="child body fixed angular velocity in N measure...

`kinematic_equations()` returns what q' equals if we define u=q', right? Basically it hard codes the result of solving the kinematic equations for q', so I think we should use that...

> Since they leave the qdot's in the equations. They are in the equations but always in the form +/- qdot, so easily removed.