I need to render/simulate a third-party model, which contains a bone structure with constraints. However, I'm not too familiar with how to use these constraints as the most I've done with ragdolls so far is some simple simulations where I just force bones to stay within a certain angle from the original position.

Here, however, I get three quats and a vector per bone (as constraint parameters) and I'm really not sure what to do with them. One quat seems to always be (-0.5 -0.5 -0.5 0.5). There also seems to be a number that could indicate the type of connection, but I guess I can figure out the rest once I know what to do with the data I have so far.

Everything I've managed to find is on various engine-related forums, which I can't really use since the code will eventually have to go into a javascript app.

Difficult to say. I would *speculate* that two quats define the local joint frames on each body. The remaining quat might define the relative orientation in the rest pose and the vector might define the swing axis.

If you provide more information, eg what engine I might be able to help better.

I meant the quats are specific to the constraints, i.e. the rest pose is defined separately.

Actually when I said quats I actually meant PR transforms, and the three seem to determine the same frame in world, parent and local coordinates. And the vector seems to be the cone axis or something (I can't find any models with hinge joints so not sure). The actual angles are also there.

I'm pretty sure they use a proprietary engine though.