Practical Workflow for Reinforcement Learning Flight Controller

I am currently trying to implement Reinforcement Learning into the flight controller and have a few questions.
I have already completed tutorials on RL using simple libraries such as Stable Baselines 3.
However, I am wondering how to make it possible for an edge device like a Companion Computer, which is used with the PX4 Autopilot system.

I have found that there are lightweight libraries available for use with edge devices, such as TF Lite and PyTorch Mobile. However, there are only a few works published online regarding flight controllers enhanced by RL.

Can someone advise on what developers normally use to implement RL for flight control? This includes training the model on a high-performance computer and deploying it on the companion computer.

hi, have you made any progress ,I am also doing a similar project?