Reasoning starting from the ground

[zoechann]

How do I deploy the Instinct-BeyondMimic-Plane-G1-v0 task I trained at InstinctLab onto a g1 using instinct_onboard? My trained policy involves getting up from the ground (initial state is lying flat on the ground). Does it support reasoning starting from the ground?

[ZiwenZhuang]

Hi,

One thing to notice that, BeyondMimic's demo requires odometry system to provide linear velocity estimation. However, we didn't implement that.

You may refer to the TrackerAgent and write your own main script with the state machine.

Best,