Testing a trained policy

[Fatimah-Alahmed]

Hello,
Thank you for the great work! I successfully trained the quadrotor using the provided interface, and the training process went smoothly. I was wondering if it is possible to test the trained policy in a simulation environment instead of deploying it directly on a real drone. Is there any implementation for that?

[mzahana]

Thanks @jonas-eschmann for the great work!
I would also like to know how to test the trained policies in simulation before deploying on actual drones.

Thanks.

[jonas-eschmann]

Hi @Fatimah-Alahmed and @mzahana,

can you be a bit more specific? In general, it is trained in simulation so you can watch the trained policies as e.g. described here.

For PX4 I used to test it with the HITL cf. here:
https://github.com/rl-tools/px4

but there might be some issues as described here:
rl-tools/rl-tools#11

I hope I'll find some time soon to test it and make it work with the latest version of RLtools again (v2)