Hi everyone,

I've recently run into an issue with Agent Layer versions in my crowd simulations. I've noticed that when using the new Agent Layer (2.0) compared to an older version of Agent Layer, I'm getting different results even though I'm using the exact same model and the same bind points.

Everything works as expected with the older version, but after switching to Agent Layer 2.0, the same setup produces different behaviors/outcomes. This has me quite puzzled. 😖

I've prepared a comparison video and the relevant hip file to demonstrate the problem more clearly.

Has anyone else encountered a similar situation, or could anyone offer some insight into why this might be happening? Any suggestions or ideas would be greatly appreciated! 🙏

I will attach the comparison video and hip file shortly.

Thanks, everyone!

The second version isn't actually attaching the prop to the joint, since the joint name `mixamorig_RightHand` is invalid (should be mixamorig:RightHand)

The main difference, though, is that the old version had a "Keep Position when Attaching" parameter which fills out the Shape Transform parameters to preserve the original world-space transform when parenting the shape to the agent's joint. The 2.0 version doesn't do this, in favour of more procedural approaches like importing rigid shape bindings from the KineFX Capture Packed Geometry SOP (https://vimeo.com/681436612)