articulations-robot-demo VS RAWSim-O

I'm trying to modify this demo, which uses the UR3 robotic arm with ML-Agents (Release 1.2.0 in Unity 2020.3.46f1). I want to add imitation learning, but there seems to be an issue with the manual controls. All joints work fine except Wrist01 (which should use the O/P keys). Wrist01 moves on its own and ignores O/P key input (in the clip I rotate the base at 00:06, but I never touch the wrist controls). Something's likely wrong with one of these files: "RobotController" (called by the parent UnityRobotics GameObject), "ArticulationJointController" (called by the child Wrist01 GameObject), or "RobotManualInput" (called by the ManualInput GameObject), but I can't for the life of me identify where the error is. Anyone know what's going wrong?

For reference i am using https://github.com/Unity-Technologies/articulations-robot-demo Robotiq Hand-E Gripper. It can only go up and down but I'm trying to make it go left and right as well.

I got this post forwarded by a friend. I hope this is still alive. I did some research on RMFS back in the day. I also shared basically all my code here: https://github.com/merschformann/RAWSim-O Maybe it's useful to you in some way (apologies for some wild-growth in the code base in advance; it grew quickly and had to satisfy many ideas). Personally, I was interested in more accurate results (not too many abstractions), which is why I started with a detailed simulation model. Then, I focused on problems that arose from that approach (path planning, basic control logic, etc.). I later worked for a supplier of an RMFS and was able to verify the accuracy of my model to some extent (it came quite close - depending on the parameters of course). However, there are certainly different angles from which to approach RMFS research-wise. It also depends a bit on whether specific methodologies are expected from you (you mentioned Queuing Networks, etc.). I can also highly recommend finding someone for collaboration (it gave me a lot, not only regarding research but also personally - I met my best research buddy at a conference). I may be a bit out of touch with the state-of-the-art as I moved to new applications. But if it helps, feel free to ping me for some discussion. (I had quite some people contacting me about this, and I'm always happy to help [as time allows it and as helpful as I can be ;) ])