Fail to train the environment with RL

[jc-bao]

The quadrotor seems to hit the boundary rather than try to stabilize itself.

Here is the performance of the lower-level controller:

[jc-bao]

Here is the random exploration render:	render	plot

[jc-bao]

Action

resolve the lower-level controller error first.

[jc-bao]

🤗 Summary

This problem is resolved by changing the rope's elastic coefficient. The large k lead to unstable simulation.