cypypccpy / dynamic_handover Goto Github PK

Hi, this work is very similar to the throwing work in the paper "DexPBT: Scaling up Dexterous Manipulation for Hand-Arm Systems with Population Based Training", but It's a great work with a different approach and used on a real robot!
My most recent research has been on tasks involving dexterous hand grasping. The current progress is that I have completed the training for grasping, but I will only be using the Ros communication node to map the motion from Isaac gym to a physical robot for open-loop sim2real. if I use the state feedback from the real robot, it will be very ineffective maybe due to the communication delays. May I ask what is your specific sim2real implementation like? Is the inference of the neural network done outside of the Isaac program?

Looking forward to your reply, thanks!
DannyChen