1
$\begingroup$

I completed the Udacity nanodegree in Deep Learning but found the final project to be extremely difficult since the only visual feedback were from plots via matplotlib (rotor speeds, x/y/z coordinates, x/y/z velocity, etc). Is there any way to render the learning per episode as done in openai gym?

A quick search for Python game engines only confused me more in terms of which one to use.

Here is a link to the final project, ie. fly a quadcopter using DDPG algorithm: https://github.com/udacity/RL-Quadcopter-2

It has a physics_sim.py, task.py, agent.py and a main ipynb notebook.

$\endgroup$

1 Answer 1

1
$\begingroup$

I tried contacting a few drone simulator makers including DRL (Drone Racing League) and Velocidrone but the latter does not have API for Python and DRL has not responded.

Therefore, I looked at Unity ML Agents and saw that it's possible to create a simulation environment using free drone assets:

https://unity3d.com/how-to/unity-machine-learning-agents

https://assetstore.unity.com/packages/tools/physics/free-pack-117641

And they have a Python API:

https://github.com/Unity-Technologies/ml-agents/blob/master/docs/Python-API.md

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.