I can only share experience I did with the A6 physics engine in combination with bones animation.

If you plan to use a rigged model with bones to use for your physics simulation: forget it or put all your effort into it. The problems are everywhere you look at. The most disgusting habit the A6 bones have is that the way they are saved in MED, they all have a local default orientation of (0,0,0). That means: no matter which absolute angle they would have - if you set their pan parameter for instance =10, they are pointing in that direction, the pan is only added to their local orientation. So you must know for each bone which direction he has. Basically, I aligned all bones for my ragdoll to the axis and defined template rotation for each bone, so that I can easily define for each bone update which axis he is aligned to.

So.. when you update each frame the bones you have to reset each bone, so that when you apply a spacial vector to move to the tied physics hinge, you only need to apply the difference vector. When applying rotation you just need to reset the orientation so that the bones "look" into absolute (0,0,0) direction. Then you can simply calculate the angle from the tied hinge entity to the child hinge and apply that angle to the tied bone. Just math calculations.

One important thing: dont forget to move the origin of the bones model BEFORE you update the bones, because when you dont move it, sometimes the modelshape enters the camera or so and then it gets invisible.. strange.

I hope this theory helps you to setup your ragdoll. I did it one month (or two) ago for REVOgames and it worked well. Though.. there are some serious problems with calculating bone rotation, I strongly advice to use a ragdoll simulation that takes multiple meshes to simulate the ragdoll body parts, because you would avoid some rotation issues I came along with. Plus, you can seperate body parts from the ragdoll, if you wish.

Cheers!
Christian